EvenMore-pytorch-models-Accuracies.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Even More pretrained models in PyTorch\n",
    "\n",
    "####  https://github.com/Cadene/pretrained-models.pytorch\n",
    "\n",
    "Can we correlate power law exponents with accuracy ?\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-08-27T08:06:08.823252Z",
     "start_time": "2018-08-27T08:06:06.545222Z"
    }
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision.models as models\n",
    "import torchvision.transforms as transforms\n",
    "import torch.nn as nn\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import numpy as np\n",
    "import powerlaw\n",
    "\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "\n",
    "import matplotlib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-08-27T08:06:11.499813Z",
     "start_time": "2018-08-27T08:06:08.826030Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "importing Jupyter notebook from RMT_Util.ipynb\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6\n",
      "  return f(*args, **kwds)\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3.6.5 |Anaconda, Inc.| (default, Apr 26 2018, 08:42:37) \n",
      "[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)]\n",
      "numpy version 1.14.3\n",
      "scipy version 1.1.0\n",
      "tensforflow version 1.10.1\n",
      "keras version 2.2.2\n",
      "sklearn version 0.19.1\n",
      "powerlaw version 1.4.3\n"
     ]
    }
   ],
   "source": [
    "import import_ipynb\n",
    "import RMT_Util"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-08-27T08:06:11.506733Z",
     "start_time": "2018-08-27T08:06:11.501899Z"
    }
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams['figure.figsize'] = [5,5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pretrainedmodels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# does not work...I made symbolic link instead\n",
    "# %env TORCH_MODEL_ZOO=\"~/work-tmp/pretrainedmodels\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 45 Available models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['fbresnet152',\n",
       " 'bninception',\n",
       " 'resnext101_32x4d',\n",
       " 'resnext101_64x4d',\n",
       " 'inceptionv4',\n",
       " 'inceptionresnetv2',\n",
       " 'alexnet',\n",
       " 'densenet121',\n",
       " 'densenet169',\n",
       " 'densenet201',\n",
       " 'densenet161',\n",
       " 'resnet18',\n",
       " 'resnet34',\n",
       " 'resnet50',\n",
       " 'resnet101',\n",
       " 'resnet152',\n",
       " 'inceptionv3',\n",
       " 'squeezenet1_0',\n",
       " 'squeezenet1_1',\n",
       " 'vgg11',\n",
       " 'vgg11_bn',\n",
       " 'vgg13',\n",
       " 'vgg13_bn',\n",
       " 'vgg16',\n",
       " 'vgg16_bn',\n",
       " 'vgg19_bn',\n",
       " 'vgg19',\n",
       " 'nasnetamobile',\n",
       " 'nasnetalarge',\n",
       " 'dpn68',\n",
       " 'dpn68b',\n",
       " 'dpn92',\n",
       " 'dpn98',\n",
       " 'dpn131',\n",
       " 'dpn107',\n",
       " 'xception',\n",
       " 'senet154',\n",
       " 'se_resnet50',\n",
       " 'se_resnet101',\n",
       " 'se_resnet152',\n",
       " 'se_resnext50_32x4d',\n",
       " 'se_resnext101_32x4d',\n",
       " 'cafferesnet101',\n",
       " 'pnasnet5large',\n",
       " 'polynet']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pretrainedmodels.model_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "62"
      ]
     },
     "execution_count": 191,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracies = pd.read_csv('accuracies', delimiter=\"\\t\")\n",
    "len(accuracies)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### TODO: remove duplcates when can use our ourporting\n",
    "\n",
    "1. get all ourporting\n",
    "2. get the other pytorch models\n",
    "\n",
    "I think this is enough"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Tensorflow', 'ourporting', 'Caffe', 'Torch7', 'Keras', 'Pytorch'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 192,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracies['platform'].unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24,                   model    platform    acc1    acc5\n",
       " 1       PNASNet-5-Large  ourporting  82.736  95.992\n",
       " 3        NASNet-A-Large  ourporting  82.566  96.086\n",
       " 5              SENet154  ourporting  81.304  95.498\n",
       " 7               PolyNet  ourporting  81.002  95.624\n",
       " 10  SE-ResNeXt101_32x4d  ourporting  80.236  95.028\n",
       " 12    InceptionResNetV2  ourporting  80.170  95.234\n",
       " 13          InceptionV4  ourporting  80.062  94.926\n",
       " 14    DualPathNet107_5k  ourporting  79.746  94.684\n",
       " 16       DualPathNet131  ourporting  79.432  94.574\n",
       " 17     DualPathNet92_5k  ourporting  79.400  94.620\n",
       " 18        DualPathNet98  ourporting  79.224  94.488\n",
       " 19   SE-ResNeXt50_32x4d  ourporting  79.076  94.434\n",
       " 22     ResNeXt101_64x4d  ourporting  78.956  94.252\n",
       " 23             Xception  ourporting  78.888  94.292\n",
       " 26         SE-ResNet152  ourporting  78.658  94.374\n",
       " 28         SE-ResNet101  ourporting  78.396  94.258\n",
       " 30     ResNeXt101_32x4d  ourporting  78.188  93.886\n",
       " 33          SE-ResNet50  ourporting  77.636  93.752\n",
       " 36          FBResNet152  ourporting  77.386  93.594\n",
       " 39    DualPathNet68b_5k  ourporting  77.034  93.590\n",
       " 41       CaffeResnet101  ourporting  76.200  92.766\n",
       " 44        DualPathNet68  ourporting  75.868  92.774\n",
       " 48      NASNet-A-Mobile  ourporting  74.080  91.740\n",
       " 50          BNInception  ourporting  73.522  91.560)"
      ]
     },
     "execution_count": 193,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "our_acc = accuracies[accuracies['platform']=='ourporting']\n",
    "len(our_acc), our_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(21,             model platform    acc1    acc5\n",
       " 27      ResNet152  Pytorch  78.428  94.110\n",
       " 34    DenseNet161  Pytorch  77.560  93.798\n",
       " 35      ResNet101  Pytorch  77.438  93.672\n",
       " 37    InceptionV3  Pytorch  77.294  93.454\n",
       " 38    DenseNet201  Pytorch  77.152  93.548\n",
       " 42    DenseNet169  Pytorch  76.026  92.992\n",
       " 43       ResNet50  Pytorch  76.002  92.980\n",
       " 45    DenseNet121  Pytorch  74.646  92.136\n",
       " 46       VGG19_BN  Pytorch  74.266  92.066\n",
       " 49       ResNet34  Pytorch  73.554  91.456\n",
       " 51       VGG16_BN  Pytorch  73.518  91.608\n",
       " 52          VGG19  Pytorch  72.080  90.822\n",
       " 53          VGG16  Pytorch  71.636  90.354\n",
       " 54       VGG13_BN  Pytorch  71.508  90.494\n",
       " 55       VGG11_BN  Pytorch  70.452  89.818\n",
       " 56       ResNet18  Pytorch  70.142  89.274\n",
       " 57          VGG13  Pytorch  69.662  89.264\n",
       " 58          VGG11  Pytorch  68.970  88.746\n",
       " 59  SqueezeNet1_1  Pytorch  58.250  80.800\n",
       " 60  SqueezeNet1_0  Pytorch  58.108  80.428\n",
       " 61        Alexnet  Pytorch  56.432  79.194)"
      ]
     },
     "execution_count": 194,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "py_acc = accuracies[accuracies['platform']=='Pytorch']\n",
    "len(py_acc), py_acc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- merge dataframes\n",
    "- get model names\n",
    "- load pretrained\n",
    "- compute alpha\n",
    "- store in data frame"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>platform</th>\n",
       "      <th>acc1</th>\n",
       "      <th>acc5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>ResNet152</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>78.428</td>\n",
       "      <td>94.110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>DenseNet161</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.560</td>\n",
       "      <td>93.798</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>ResNet101</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.438</td>\n",
       "      <td>93.672</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>InceptionV3</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.294</td>\n",
       "      <td>93.454</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>DenseNet201</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.152</td>\n",
       "      <td>93.548</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          model platform    acc1    acc5\n",
       "27    ResNet152  Pytorch  78.428  94.110\n",
       "34  DenseNet161  Pytorch  77.560  93.798\n",
       "35    ResNet101  Pytorch  77.438  93.672\n",
       "37  InceptionV3  Pytorch  77.294  93.454\n",
       "38  DenseNet201  Pytorch  77.152  93.548"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df =  pd.concat([py_acc, our_acc])\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Minimum dimension for W"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "MIN_DIM = 50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### All Linear Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-08-27T13:12:09.422260Z",
     "start_time": "2018-08-27T13:12:09.341227Z"
    }
   },
   "outputs": [],
   "source": [
    "def count_layers(model):\n",
    "    \"\"\"Loops over a pytorch model and counts all the weight matrices, linear, conv2d, and conv2d-slices\"\"\"\n",
    "    num_linear, num_conv2d, num_conv2d_slices = 0, 0, 0\n",
    "    \n",
    "    for im, m in enumerate(model.modules()):\n",
    "        #print(im,m)\n",
    "        if isinstance(m, nn.Linear):     \n",
    "            W = np.array(m.weight.data.clone().cpu())\n",
    "            if W is not None:\n",
    "                #print(im,m, W.shape)\n",
    "                num_linear += 1\n",
    "        elif isinstance(m, nn.Conv2d):\n",
    "            #print(\"conv2d\")\n",
    "            Wtensor = np.array(m.weight.data.clone().cpu())\n",
    "            #print(Wtensor.shape)\n",
    "            Wmats = RMT_Util.get_conv2D_Wmats(Wtensor)\n",
    "            #print(im,m)\n",
    "            num_conv2d += 1\n",
    "            for W in Wmats:\n",
    "                #print(W.shape)\n",
    "                if (np.min(W.shape) >= MIN_DIM):\n",
    "                    num_conv2d_slices += 1\n",
    "                \n",
    "    return num_linear, num_conv2d, num_conv2d_slices\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def iter_pytotch_layers(model, min_dim=MIN_DIM):\n",
    "    \"\"\"Iterate over a pytorch layers and  weight matrix.  \n",
    "    Supports linear, conv2d, and conv2d-slices\"\"\"\n",
    "    for im, m in enumerate(model.modules()):\n",
    "        if isinstance(m, nn.Linear):     \n",
    "            W = np.array(m.weight.data.clone().cpu())\n",
    "            if W is not None:\n",
    "                yield W\n",
    "        elif isinstance(m, nn.Conv2d):\n",
    "            Wtensor = np.array(m.weight.data.clone().cpu())\n",
    "            Wmats = RMT_Util.get_conv2D_Wmats(Wtensor)\n",
    "            for W in Wmats:\n",
    "                if (np.min(W.shape) >= min_dim):\n",
    "                    yield W"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def calc_model_alphas(model):\n",
    "    try:\n",
    "        alphas, min_svs, Qs = [], [], []\n",
    "\n",
    "        for W in tqdm(iter_pytotch_layers(model)):\n",
    "            M, N = np.min(W.shape), np.max(W.shape)\n",
    "            Q = N/M\n",
    "            Qs.append(Q)\n",
    "\n",
    "            _, sv, _ = np.linalg.svd(W)\n",
    "            min_svs.append(np.min(sv))\n",
    "\n",
    "            evals = sv*sv\n",
    "            fit = powerlaw.Fit(evals, xmax=np.max(evals), verbose=False)\n",
    "            alphas.append(fit.alpha)\n",
    "\n",
    "    except:\n",
    "        print(\"skipping \", model_name)\n",
    "        pass\n",
    "    \n",
    "    return alphas, min_svs, Qs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "VGG(\n",
       "  (linear0): Linear(in_features=25088, out_features=4096, bias=True)\n",
       "  (relu0): ReLU(inplace)\n",
       "  (dropout0): Dropout(p=0.5)\n",
       "  (linear1): Linear(in_features=4096, out_features=4096, bias=True)\n",
       "  (relu1): ReLU(inplace)\n",
       "  (dropout1): Dropout(p=0.5)\n",
       "  (last_linear): Linear(in_features=4096, out_features=1000, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_name = \"vgg16\"\n",
    "model = pretrainedmodels.__dict__[model_name](num_classes=1000, pretrained='imagenet')\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6de3c562389340448a9136ecb334b5a8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:700: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  (Theoretical_CDF * (1 - Theoretical_CDF))\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:700: RuntimeWarning: invalid value encountered in true_divide\n",
      "  (Theoretical_CDF * (1 - Theoretical_CDF))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "alphas, min_svs, Qs = calc_model_alphas(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[2.2964333532786694, 2.18367837003559, 2.093936912920813]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alphas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4b7585b0a60b4b01a64f64765b4d6e3b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:700: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  (Theoretical_CDF * (1 - Theoretical_CDF))\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:700: RuntimeWarning: invalid value encountered in true_divide\n",
      "  (Theoretical_CDF * (1 - Theoretical_CDF))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/torchvision/models/densenet.py:212: UserWarning: nn.init.kaiming_normal is now deprecated in favor of nn.init.kaiming_normal_.\n",
      "  nn.init.kaiming_normal(m.weight.data)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ea28110e77bd479aaf37ea28b19dbe64",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fc0406aed093485186a93afc8eee518f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3fa4793ed84240789858fd94d00f901d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bbe4b6e155bf4aec9a0458db31e22fcd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "228543d984aa4ebfb23a9551551a3440",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1986d154f3984efda336b2cf41f6c0c6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d93c8e03f9324da382a4894e59863625",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02d606f9517d45ba993083ccf2ddbb53",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "38032023abea476da26cdce1f91ce0aa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6df7f3fcce8346b4aa4109c8bd6b2a3d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7f8f19338fab4313b50cde4a2dcaf6b0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "966dbfa9d41d469d9b30e1668d8c35c0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b171822f97034fdb8d15c5f398c2917b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/torchvision/models/squeezenet.py:94: UserWarning: nn.init.kaiming_uniform is now deprecated in favor of nn.init.kaiming_uniform_.\n",
      "  init.kaiming_uniform(m.weight.data)\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/torchvision/models/squeezenet.py:92: UserWarning: nn.init.normal is now deprecated in favor of nn.init.normal_.\n",
      "  init.normal(m.weight.data, mean=0.0, std=0.01)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0aba18bf34ac4c12be49f79db6564593",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:1178: RuntimeWarning: overflow encountered in double_scalars\n",
      "  return (self.alpha-1) * self.xmin**(self.alpha-1)\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:825: RuntimeWarning: invalid value encountered in multiply\n",
      "  likelihoods = f*C\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/scipy/optimize/optimize.py:563: RuntimeWarning: invalid value encountered in subtract\n",
      "  numpy.max(numpy.abs(fsim[0] - fsim[1:])) <= fatol):\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eb0af7e30c7d4b3faf75e82280b36b3b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ce6da3cd27064f068768e27a470023ed",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "29730b3fcb614686adc15a7c1c31557b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1add1c0314b74977901d15412becfbbf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "196a7064a7f14136840cbd71e4a78b37",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ecc759ee18ba43c3a9dbad327d91c48c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/powerlaw.py:1174: RuntimeWarning: overflow encountered in power\n",
      "  return x**-self.alpha\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2405ea1c51fe4390a0873dd1f7d63edb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e5ee97bfff3b4208bed7d1605bb118bb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "84e6c2a47e9a4d3f93a2d656b951d096",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c8d726a7f8d44dc0a2d9a1aa7a5c8531",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eec0ae5eff7a4634a7adcd72b72f2f5d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "53ff4992dbba43b189772225d233d628",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "72cc527c48b4438481a53781e47b8097",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "alphas_4model = {}\n",
    "counts_4model = {}\n",
    "models = [m.lower().replace('-','') for m in df['model'].values]\n",
    "\n",
    "for model_name in models:\n",
    "    if(model_name in pretrainedmodels.model_names):\n",
    "        \n",
    "        try:\n",
    "            model = pretrainedmodels.__dict__[model_name](num_classes=1000, pretrained='imagenet')\n",
    "            model.eval()\n",
    "            \n",
    "            alphas, min_svs, Qs = calc_model_alphas(model)\n",
    "            counts_4model[model_name] = count_layers(model)\n",
    "            alphas_4model[model_name] = [alphas, min_svs, Qs]\n",
    "\n",
    "        except:\n",
    "            pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>platform</th>\n",
       "      <th>acc1</th>\n",
       "      <th>acc5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>ResNet152</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>78.428</td>\n",
       "      <td>94.110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>DenseNet161</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.560</td>\n",
       "      <td>93.798</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>ResNet101</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.438</td>\n",
       "      <td>93.672</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>InceptionV3</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.294</td>\n",
       "      <td>93.454</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>DenseNet201</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.152</td>\n",
       "      <td>93.548</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          model platform    acc1    acc5\n",
       "27    ResNet152  Pytorch  78.428  94.110\n",
       "34  DenseNet161  Pytorch  77.560  93.798\n",
       "35    ResNet101  Pytorch  77.438  93.672\n",
       "37  InceptionV3  Pytorch  77.294  93.454\n",
       "38  DenseNet201  Pytorch  77.152  93.548"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>platform</th>\n",
       "      <th>acc1</th>\n",
       "      <th>acc5</th>\n",
       "      <th>modelname</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>ResNet152</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>78.428</td>\n",
       "      <td>94.110</td>\n",
       "      <td>resnet152</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>DenseNet161</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.560</td>\n",
       "      <td>93.798</td>\n",
       "      <td>densenet161</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>ResNet101</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.438</td>\n",
       "      <td>93.672</td>\n",
       "      <td>resnet101</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          model platform    acc1    acc5    modelname\n",
       "27    ResNet152  Pytorch  78.428  94.110    resnet152\n",
       "34  DenseNet161  Pytorch  77.560  93.798  densenet161\n",
       "35    ResNet101  Pytorch  77.438  93.672    resnet101"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['modelname'] = df['model'].apply(lambda x: x.lower().replace('-',''))\n",
    "df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {},
   "outputs": [],
   "source": [
    "avg_alphas = {}\n",
    "avg_alphas_m1 = {}\n",
    "len_alphas = {}\n",
    "for model_name in models:\n",
    "    avg_alphas[model_name] = 0.0\n",
    "    len_alphas[model_name] = 0.0\n",
    "    avg_alphas_m1[model_name] = 0.0\n",
    "    \n",
    "    if(model_name in alphas_4model):\n",
    "        avg_alphas[model_name] = np.average(alphas_4model[model_name][0])\n",
    "        avg_alphas_m1[model_name] = np.average(alphas_4model[model_name][0][:-1])\n",
    "\n",
    "        len_alphas[model_name] = len(alphas_4model[model_name][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['avg_alpha'] = df['modelname'].apply(lambda x: avg_alphas[x] )\n",
    "df['avg_alpha_m1'] = df['modelname'].apply(lambda x: avg_alphas_m1[x] )\n",
    "\n",
    "df['num_layers'] = df['modelname'].apply(lambda x: int(len_alphas[x]) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>platform</th>\n",
       "      <th>acc1</th>\n",
       "      <th>acc5</th>\n",
       "      <th>modelname</th>\n",
       "      <th>avg_alpha</th>\n",
       "      <th>num_layers</th>\n",
       "      <th>avg_alpha_m1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>61</th>\n",
       "      <td>Alexnet</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>56.432</td>\n",
       "      <td>79.194</td>\n",
       "      <td>alexnet</td>\n",
       "      <td>2.800691</td>\n",
       "      <td>55</td>\n",
       "      <td>2.796633</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>BNInception</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>73.522</td>\n",
       "      <td>91.560</td>\n",
       "      <td>bninception</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>CaffeResnet101</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>76.200</td>\n",
       "      <td>92.766</td>\n",
       "      <td>cafferesnet101</td>\n",
       "      <td>2.997881</td>\n",
       "      <td>368</td>\n",
       "      <td>2.995803</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>DenseNet121</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>74.646</td>\n",
       "      <td>92.136</td>\n",
       "      <td>densenet121</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>DenseNet161</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.560</td>\n",
       "      <td>93.798</td>\n",
       "      <td>densenet161</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>DenseNet169</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>76.026</td>\n",
       "      <td>92.992</td>\n",
       "      <td>densenet169</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>DenseNet201</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.152</td>\n",
       "      <td>93.548</td>\n",
       "      <td>densenet201</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>DualPathNet107_5k</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>79.746</td>\n",
       "      <td>94.684</td>\n",
       "      <td>dualpathnet107_5k</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>DualPathNet131</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>79.432</td>\n",
       "      <td>94.574</td>\n",
       "      <td>dualpathnet131</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>DualPathNet68</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>75.868</td>\n",
       "      <td>92.774</td>\n",
       "      <td>dualpathnet68</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>DualPathNet68b_5k</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>77.034</td>\n",
       "      <td>93.590</td>\n",
       "      <td>dualpathnet68b_5k</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>DualPathNet92_5k</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>79.400</td>\n",
       "      <td>94.620</td>\n",
       "      <td>dualpathnet92_5k</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>DualPathNet98</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>79.224</td>\n",
       "      <td>94.488</td>\n",
       "      <td>dualpathnet98</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>FBResNet152</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>77.386</td>\n",
       "      <td>93.594</td>\n",
       "      <td>fbresnet152</td>\n",
       "      <td>2.706008</td>\n",
       "      <td>555</td>\n",
       "      <td>2.704413</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>InceptionResNetV2</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>80.170</td>\n",
       "      <td>95.234</td>\n",
       "      <td>inceptionresnetv2</td>\n",
       "      <td>3.356050</td>\n",
       "      <td>540</td>\n",
       "      <td>3.354540</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>InceptionV3</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.294</td>\n",
       "      <td>93.454</td>\n",
       "      <td>inceptionv3</td>\n",
       "      <td>3.414726</td>\n",
       "      <td>396</td>\n",
       "      <td>3.412019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>InceptionV4</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>80.062</td>\n",
       "      <td>94.926</td>\n",
       "      <td>inceptionv4</td>\n",
       "      <td>2.853903</td>\n",
       "      <td>627</td>\n",
       "      <td>2.851791</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>NASNet-A-Large</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>82.566</td>\n",
       "      <td>96.086</td>\n",
       "      <td>nasnetalarge</td>\n",
       "      <td>3.756205</td>\n",
       "      <td>255</td>\n",
       "      <td>3.757384</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>NASNet-A-Mobile</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>74.080</td>\n",
       "      <td>91.740</td>\n",
       "      <td>nasnetamobile</td>\n",
       "      <td>4.500251</td>\n",
       "      <td>121</td>\n",
       "      <td>4.511308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>PNASNet-5-Large</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>82.736</td>\n",
       "      <td>95.992</td>\n",
       "      <td>pnasnet5large</td>\n",
       "      <td>3.894849</td>\n",
       "      <td>205</td>\n",
       "      <td>3.897871</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>PolyNet</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>81.002</td>\n",
       "      <td>95.624</td>\n",
       "      <td>polynet</td>\n",
       "      <td>2.889384</td>\n",
       "      <td>815</td>\n",
       "      <td>2.888961</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>ResNeXt101_32x4d</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>78.188</td>\n",
       "      <td>93.886</td>\n",
       "      <td>resnext101_32x4d</td>\n",
       "      <td>4.430921</td>\n",
       "      <td>71</td>\n",
       "      <td>4.445087</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>ResNeXt101_64x4d</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>78.956</td>\n",
       "      <td>94.252</td>\n",
       "      <td>resnext101_64x4d</td>\n",
       "      <td>3.762263</td>\n",
       "      <td>71</td>\n",
       "      <td>3.768079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>ResNet101</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>77.438</td>\n",
       "      <td>93.672</td>\n",
       "      <td>resnet101</td>\n",
       "      <td>2.966080</td>\n",
       "      <td>368</td>\n",
       "      <td>2.964438</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>ResNet152</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>78.428</td>\n",
       "      <td>94.110</td>\n",
       "      <td>resnet152</td>\n",
       "      <td>2.724560</td>\n",
       "      <td>555</td>\n",
       "      <td>2.723115</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56</th>\n",
       "      <td>ResNet18</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>70.142</td>\n",
       "      <td>89.274</td>\n",
       "      <td>resnet18</td>\n",
       "      <td>3.353278</td>\n",
       "      <td>148</td>\n",
       "      <td>3.353343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>ResNet34</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>73.554</td>\n",
       "      <td>91.456</td>\n",
       "      <td>resnet34</td>\n",
       "      <td>3.329474</td>\n",
       "      <td>292</td>\n",
       "      <td>3.329268</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>ResNet50</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>76.002</td>\n",
       "      <td>92.980</td>\n",
       "      <td>resnet50</td>\n",
       "      <td>3.398341</td>\n",
       "      <td>181</td>\n",
       "      <td>3.397541</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>SE-ResNeXt101_32x4d</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>80.236</td>\n",
       "      <td>95.028</td>\n",
       "      <td>seresnext101_32x4d</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>SE-ResNeXt50_32x4d</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>79.076</td>\n",
       "      <td>94.434</td>\n",
       "      <td>seresnext50_32x4d</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>SE-ResNet101</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>78.396</td>\n",
       "      <td>94.258</td>\n",
       "      <td>seresnet101</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>SE-ResNet152</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>78.658</td>\n",
       "      <td>94.374</td>\n",
       "      <td>seresnet152</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>SE-ResNet50</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>77.636</td>\n",
       "      <td>93.752</td>\n",
       "      <td>seresnet50</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>SENet154</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>81.304</td>\n",
       "      <td>95.498</td>\n",
       "      <td>senet154</td>\n",
       "      <td>2.723808</td>\n",
       "      <td>225</td>\n",
       "      <td>2.716392</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>SqueezeNet1_0</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>58.108</td>\n",
       "      <td>80.428</td>\n",
       "      <td>squeezenet1_0</td>\n",
       "      <td>3.604989</td>\n",
       "      <td>23</td>\n",
       "      <td>3.646235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59</th>\n",
       "      <td>SqueezeNet1_1</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>58.250</td>\n",
       "      <td>80.800</td>\n",
       "      <td>squeezenet1_1</td>\n",
       "      <td>3.475992</td>\n",
       "      <td>23</td>\n",
       "      <td>3.506683</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58</th>\n",
       "      <td>VGG11</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>68.970</td>\n",
       "      <td>88.746</td>\n",
       "      <td>vgg11</td>\n",
       "      <td>2.439411</td>\n",
       "      <td>3</td>\n",
       "      <td>2.246289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>VGG11_BN</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>70.452</td>\n",
       "      <td>89.818</td>\n",
       "      <td>vgg11_bn</td>\n",
       "      <td>2.338732</td>\n",
       "      <td>3</td>\n",
       "      <td>2.010740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>57</th>\n",
       "      <td>VGG13</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>69.662</td>\n",
       "      <td>89.264</td>\n",
       "      <td>vgg13</td>\n",
       "      <td>2.415459</td>\n",
       "      <td>3</td>\n",
       "      <td>2.262122</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54</th>\n",
       "      <td>VGG13_BN</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>71.508</td>\n",
       "      <td>90.494</td>\n",
       "      <td>vgg13_bn</td>\n",
       "      <td>2.339144</td>\n",
       "      <td>3</td>\n",
       "      <td>2.011003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53</th>\n",
       "      <td>VGG16</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>71.636</td>\n",
       "      <td>90.354</td>\n",
       "      <td>vgg16</td>\n",
       "      <td>2.191350</td>\n",
       "      <td>3</td>\n",
       "      <td>2.240056</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51</th>\n",
       "      <td>VGG16_BN</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>73.518</td>\n",
       "      <td>91.608</td>\n",
       "      <td>vgg16_bn</td>\n",
       "      <td>2.350616</td>\n",
       "      <td>3</td>\n",
       "      <td>2.010206</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>VGG19</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>72.080</td>\n",
       "      <td>90.822</td>\n",
       "      <td>vgg19</td>\n",
       "      <td>2.176967</td>\n",
       "      <td>3</td>\n",
       "      <td>2.229711</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>VGG19_BN</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>74.266</td>\n",
       "      <td>92.066</td>\n",
       "      <td>vgg19_bn</td>\n",
       "      <td>2.351588</td>\n",
       "      <td>3</td>\n",
       "      <td>2.011790</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>Xception</td>\n",
       "      <td>ourporting</td>\n",
       "      <td>78.888</td>\n",
       "      <td>94.292</td>\n",
       "      <td>xception</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  model    platform    acc1    acc5           modelname  \\\n",
       "61              Alexnet     Pytorch  56.432  79.194             alexnet   \n",
       "50          BNInception  ourporting  73.522  91.560         bninception   \n",
       "41       CaffeResnet101  ourporting  76.200  92.766      cafferesnet101   \n",
       "45          DenseNet121     Pytorch  74.646  92.136         densenet121   \n",
       "34          DenseNet161     Pytorch  77.560  93.798         densenet161   \n",
       "42          DenseNet169     Pytorch  76.026  92.992         densenet169   \n",
       "38          DenseNet201     Pytorch  77.152  93.548         densenet201   \n",
       "14    DualPathNet107_5k  ourporting  79.746  94.684   dualpathnet107_5k   \n",
       "16       DualPathNet131  ourporting  79.432  94.574      dualpathnet131   \n",
       "44        DualPathNet68  ourporting  75.868  92.774       dualpathnet68   \n",
       "39    DualPathNet68b_5k  ourporting  77.034  93.590   dualpathnet68b_5k   \n",
       "17     DualPathNet92_5k  ourporting  79.400  94.620    dualpathnet92_5k   \n",
       "18        DualPathNet98  ourporting  79.224  94.488       dualpathnet98   \n",
       "36          FBResNet152  ourporting  77.386  93.594         fbresnet152   \n",
       "12    InceptionResNetV2  ourporting  80.170  95.234   inceptionresnetv2   \n",
       "37          InceptionV3     Pytorch  77.294  93.454         inceptionv3   \n",
       "13          InceptionV4  ourporting  80.062  94.926         inceptionv4   \n",
       "3        NASNet-A-Large  ourporting  82.566  96.086        nasnetalarge   \n",
       "48      NASNet-A-Mobile  ourporting  74.080  91.740       nasnetamobile   \n",
       "1       PNASNet-5-Large  ourporting  82.736  95.992       pnasnet5large   \n",
       "7               PolyNet  ourporting  81.002  95.624             polynet   \n",
       "30     ResNeXt101_32x4d  ourporting  78.188  93.886    resnext101_32x4d   \n",
       "22     ResNeXt101_64x4d  ourporting  78.956  94.252    resnext101_64x4d   \n",
       "35            ResNet101     Pytorch  77.438  93.672           resnet101   \n",
       "27            ResNet152     Pytorch  78.428  94.110           resnet152   \n",
       "56             ResNet18     Pytorch  70.142  89.274            resnet18   \n",
       "49             ResNet34     Pytorch  73.554  91.456            resnet34   \n",
       "43             ResNet50     Pytorch  76.002  92.980            resnet50   \n",
       "10  SE-ResNeXt101_32x4d  ourporting  80.236  95.028  seresnext101_32x4d   \n",
       "19   SE-ResNeXt50_32x4d  ourporting  79.076  94.434   seresnext50_32x4d   \n",
       "28         SE-ResNet101  ourporting  78.396  94.258         seresnet101   \n",
       "26         SE-ResNet152  ourporting  78.658  94.374         seresnet152   \n",
       "33          SE-ResNet50  ourporting  77.636  93.752          seresnet50   \n",
       "5              SENet154  ourporting  81.304  95.498            senet154   \n",
       "60        SqueezeNet1_0     Pytorch  58.108  80.428       squeezenet1_0   \n",
       "59        SqueezeNet1_1     Pytorch  58.250  80.800       squeezenet1_1   \n",
       "58                VGG11     Pytorch  68.970  88.746               vgg11   \n",
       "55             VGG11_BN     Pytorch  70.452  89.818            vgg11_bn   \n",
       "57                VGG13     Pytorch  69.662  89.264               vgg13   \n",
       "54             VGG13_BN     Pytorch  71.508  90.494            vgg13_bn   \n",
       "53                VGG16     Pytorch  71.636  90.354               vgg16   \n",
       "51             VGG16_BN     Pytorch  73.518  91.608            vgg16_bn   \n",
       "52                VGG19     Pytorch  72.080  90.822               vgg19   \n",
       "46             VGG19_BN     Pytorch  74.266  92.066            vgg19_bn   \n",
       "23             Xception  ourporting  78.888  94.292            xception   \n",
       "\n",
       "    avg_alpha  num_layers  avg_alpha_m1  \n",
       "61   2.800691          55      2.796633  \n",
       "50   0.000000           0      0.000000  \n",
       "41   2.997881         368      2.995803  \n",
       "45   0.000000           0      0.000000  \n",
       "34   0.000000           0      0.000000  \n",
       "42   0.000000           0      0.000000  \n",
       "38   0.000000           0      0.000000  \n",
       "14   0.000000           0      0.000000  \n",
       "16   0.000000           0      0.000000  \n",
       "44   0.000000           0      0.000000  \n",
       "39   0.000000           0      0.000000  \n",
       "17   0.000000           0      0.000000  \n",
       "18   0.000000           0      0.000000  \n",
       "36   2.706008         555      2.704413  \n",
       "12   3.356050         540      3.354540  \n",
       "37   3.414726         396      3.412019  \n",
       "13   2.853903         627      2.851791  \n",
       "3    3.756205         255      3.757384  \n",
       "48   4.500251         121      4.511308  \n",
       "1    3.894849         205      3.897871  \n",
       "7    2.889384         815      2.888961  \n",
       "30   4.430921          71      4.445087  \n",
       "22   3.762263          71      3.768079  \n",
       "35   2.966080         368      2.964438  \n",
       "27   2.724560         555      2.723115  \n",
       "56   3.353278         148      3.353343  \n",
       "49   3.329474         292      3.329268  \n",
       "43   3.398341         181      3.397541  \n",
       "10   0.000000           0      0.000000  \n",
       "19   0.000000           0      0.000000  \n",
       "28   0.000000           0      0.000000  \n",
       "26   0.000000           0      0.000000  \n",
       "33   0.000000           0      0.000000  \n",
       "5    2.723808         225      2.716392  \n",
       "60   3.604989          23      3.646235  \n",
       "59   3.475992          23      3.506683  \n",
       "58   2.439411           3      2.246289  \n",
       "55   2.338732           3      2.010740  \n",
       "57   2.415459           3      2.262122  \n",
       "54   2.339144           3      2.011003  \n",
       "53   2.191350           3      2.240056  \n",
       "51   2.350616           3      2.010206  \n",
       "52   2.176967           3      2.229711  \n",
       "46   2.351588           3      2.011790  \n",
       "23   0.000000           0      0.000000  "
      ]
     },
     "execution_count": 257,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.sort_values('model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:2: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  \n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  \"\"\"\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:8: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  \n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:11: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  # This is added back by InteractiveShellApp.init_path()\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:14: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  \n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:17: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "df_vgg11 = df[(df['model']=='VGG11')| (df['model']=='VGG11_BN')]\n",
    "df_vgg11['legend']='VGG11'\n",
    "\n",
    "df_vgg13 = df[(df['model']=='VGG13')| (df['model']=='VGG13_BN')]\n",
    "df_vgg13['legend']='VGG13'\n",
    "\n",
    "df_vgg16 = df[(df['model']=='VGG16')| (df['model']=='VGG16_BN')]\n",
    "df_vgg16['legend']='VGG16'\n",
    "\n",
    "df_vgg19 = df[(df['model']=='VGG19')| (df['model']=='VGG19_BN')]\n",
    "df_vgg19['legend']='VGG19'\n",
    "\n",
    "df_squeezenet = df[(df['model']=='SqueezeNet1_0')| (df['model']=='SqueezeNet1_1')]\n",
    "df_squeezenet['legend']='SqueezeNet'\n",
    "\n",
    "df_resnxt = df[(df['model']=='ResNeXt101_32x4d')| (df['model']=='ResNeXt101_64x4d')]\n",
    "df_resnxt['legend']='ResNeXT101'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model</th>\n",
       "      <th>platform</th>\n",
       "      <th>acc1</th>\n",
       "      <th>acc5</th>\n",
       "      <th>modelname</th>\n",
       "      <th>avg_alpha</th>\n",
       "      <th>num_layers</th>\n",
       "      <th>legend</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>VGG19_BN</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>74.266</td>\n",
       "      <td>92.066</td>\n",
       "      <td>vgg19_bn</td>\n",
       "      <td>2.351588</td>\n",
       "      <td>3</td>\n",
       "      <td>VGG19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>VGG19</td>\n",
       "      <td>Pytorch</td>\n",
       "      <td>72.080</td>\n",
       "      <td>90.822</td>\n",
       "      <td>vgg19</td>\n",
       "      <td>2.176967</td>\n",
       "      <td>3</td>\n",
       "      <td>VGG19</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       model platform    acc1    acc5 modelname  avg_alpha  num_layers legend\n",
       "46  VGG19_BN  Pytorch  74.266  92.066  vgg19_bn   2.351588           3  VGG19\n",
       "52     VGG19  Pytorch  72.080  90.822     vgg19   2.176967           3  VGG19"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_vgg19"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ResNeXT101\n",
      "VGG11\n",
      "VGG13\n",
      "SqueezeNet\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for mf in [df_resnxt, df_vgg11, df_vgg13, df_squeezenet]:\n",
    "#for mf in [df_vgg11, df_vgg13]:\n",
    "    x = mf['acc5'].values\n",
    "    y = mf['avg_alpha'].values\n",
    "    label = mf['legend'].values[0]\n",
    "    print(label)\n",
    "    plt.scatter(x,y,label=label)\n",
    "\n",
    "plt.legend()\n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha)$\"+\"\\nPretrained ImageNet Models\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([2.2964333532786694, 2.18367837003559, 2.093936912920813],\n",
       " [2.052113042865135, 1.9682982563444593, 3.0314355152101187])"
      ]
     },
     "execution_count": 239,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alphas_4model['vgg16'][0], alphas_4model['vgg16_bn'][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([2.2714986725464925, 2.1879226505760556, 2.071480227760304],\n",
       " [2.042001016106473, 1.9815782182748403, 3.031184861584034])"
      ]
     },
     "execution_count": 240,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alphas_4model['vgg19'][0], alphas_4model['vgg19_bn'][0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Redo without Final Layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ResNeXT101\n",
      "VGG11\n",
      "VGG13\n",
      "VGG16\n",
      "VGG19\n",
      "SqueezeNet\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for mf in [df_resnxt, df_vgg11, df_vgg13, df_vgg16, df_vgg19, df_squeezenet]:\n",
    "#for mf in [df_vgg11, df_vgg13]:\n",
    "    x = mf['acc5'].values\n",
    "    y = mf['avg_alpha_m1'].values\n",
    "    label = mf['legend'].values[0]\n",
    "    print(label)\n",
    "    plt.scatter(x,y,label=label)\n",
    "\n",
    "plt.legend()\n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha])$\"+\"\\nPretrained ImageNet Models\\n(average excludes final layer)\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### No gross correlation\n",
    "\n",
    "Have to be more clever than this"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1, 5)"
      ]
     },
     "execution_count": 256,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(df['acc5'].values, df['avg_alpha_m1'].values)\n",
    "plt.ylim([1,5])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ResNet\n",
    "\n",
    "Only exception is InceptionResNetv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "resnet152\n",
      "resnet101\n",
      "resnet50\n",
      "resnet34\n",
      "resnet18\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1a687e8518>"
      ]
     },
     "execution_count": 267,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x, y = [], []\n",
    "for name in df.modelname:\n",
    "    if name.startswith('resnet'):\n",
    "        print(name)\n",
    "        mf = df[df.modelname==name]\n",
    "        x = mf['acc5'].values\n",
    "        y = mf['avg_alpha'].values\n",
    "        \n",
    "        plt.scatter(x,y, label=name)\n",
    "        \n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha])$\"+\"\\nPretrained Resnet### Models\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 274,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1a6a79e7f0>"
      ]
     },
     "execution_count": 274,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x, y = [], []\n",
    "for name in df.modelname:\n",
    "    if 'resnet' in name:\n",
    "        #print(name)\n",
    "        mf = df[df.modelname==name]\n",
    "        x = mf['acc5'].values\n",
    "        y = mf['avg_alpha_m1'].values\n",
    "        if y > 0:\n",
    "            plt.scatter(x,y, label=name)\n",
    "        \n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha])$\"+\"\\nPretrained ResNet Models\\n(average excludes final layer)\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 286,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1a67b5d4a8>"
      ]
     },
     "execution_count": 286,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVEAAAFgCAYAAAACUEpKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3Xt8FPX1//HXGwImchEVrFxElCpFIIaIXFRaAQUUAS9QFFuRatVaq9JK1aoVkf4EsaDfqlUrKipWqUW8tiAKClRELgG5BFEBBYMgGCTInfP7YyZxE3LbbLLJkvN8PPbB7mc++5kzG3LymZmdMzIznHPOlU2Nyg7AOecSmSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR55yLgSdR56oQSfdJujmK/u9K2iVpTkTbfEltKiZCV5AnUZdH0ixJ30o6rLJjqeoq4rOS1Ai4Ani8tO8xs+7AdQWaHwBGlldcrnieRB0AkloAXQED+sV53UnxXF+sKvCzuhJ4y8x2xjjOa0A3SY1jD8mVxJOoy3UFMA94BhgSuUDScZKmSNosaYukh4trD5eZpB9HvH5G0qiI12sl3SppKbBD0p2SPpO0XdIKSReVFIOk4ZL+XaDf3yQ9WHDjJN0m6eUCbQ9J+r/w+a2SNoTrXyWpRxk/q5LWky5pcbief0l6KeJzOQ94r8B7kyTdFX5eWyQNlvRHSXcUFZyZ7QIWAj2L2QZXXszMH/4A+BS4HjgN2Av8KGyvCSwBxgN1gGTgrKLaI8Yz4McRr58BRkW8XgtkAMcBKcBAoAnBH/ZBwA6gcQkxNA77NQj7JQGbgNMK2b7jge+B+hFjZgGdgVbAl0CTcFkLoGW0n1Up1lMbWAfcBNQCLgb25H4uwGbg9ALrGg3MAI4A+gIrgRVAvYg+VwJzCrzv/4Bxlf3/qjo8Kj0Af1T+I0xIe4GG4etMYFj4vEv4y51U4D2FtkcsL00S/VUxMWUA/UtaF/Af4Nfh8wuAFcWMOQe4Inx+LvBZ+PzHYfI9B6hV1s+qFOv5KbABUIG+uUl0L/CTiGX1gV3AieHrY8LP9U8F1ldYEv0L8FRl/9+qDg/fnXcQ7JJON7Nvwtcv8MNu6nHAOjPbV+A9RbVH48vcJ5KukJQhKVtSNtAWaFiKdU0EfhE+/wXwXDHrewG4LHw+OHyNmX0K3AyMADZJelFSkyLGKO6zKnY9BDPtDRZmudCXEc+/BepFvO4OfGJmn4evawPbgL8Vs4256gHZpejnYuRJtJqTlAL8HPiZpI2SNgLDgFMlnUrwS968kJM/RbXn+h44POL1sYX0sTCG44F/ADcAR5tZA2AZoFKsayqQKqktwUx0UjGb+y/gbEnNgIv4IblhZi+Y2VkEu+MGjCn45lJ8ViWtJwtoKkkRfY+LeL4UODnidRPgq4jX1xAk4e3FbGOu1gSHQFwF8yTqLgT2A6cAaeGjNTCb4ATKfIJf/tGS6khKlnRmMe25MoDBkmpK6g38rJgY6hAkrs0AkoYSzERzFbkuC06ivEyQqOab2RdFrcTMNgOzgKeBNWa2MlxfK0ndw68r7QJ2hp9JtJ9VsesBPgjff0N4wqg/0DFi/LfI/zmtB9IkNZbUCfglcIyk2kVtY7g9hxEcr327uH6ufHgSdUOAp83sCzPbmPsAHgYuJ5gN9iU4bvgFwS/2IDPbX1h7xLg3hcuzw3GmFhWAma0A/kqQZL4G2gFzI5aXtK6J4XuK25XP9QLBsc8XItoOIziB8w2wkeDY458KeW+xn1WBmfJB6zGzPQQnk64i+Fx+AbwB7A67PAucH854Af4LTCc4mfTP8L0ZwLslbGM/YJaZfVVCP1cOlP/wjHOJR1JzghM8x5rZd5UdTzQkfQg8ZmZPh6//H7DJzA76mlYR73+b4Mz/fDPrETHmVWa2rILCdhE8ibqEJqkGMI7gK0W/qux4SiLpZ8Aqglnv5cBjBGffsyo1MFdmCXWliHORJNUh2P1fB/Su5HBKqxUwGagLfAYM8ASa2Hwm6pxzMfATS845FwNPos45FwNPos45FwNPoi5mkppLypFUswLGHiHp+fIet7oI655eXcq++SpvudLxJJqAwrJoO8PE9bWkpyXVLeNYVyqiKnpZhF8+rxt+KT5uJJ0t6UD4OeSWsBtaTuOapEcKtM+RdGUpx4gqIYV/LEzSjQXabw7bR5R2LBdfnkQTV18zqwukA6cDdxbsoEDMP+OKmGGWo6/Cz6E+wXXs/5DUqhzG3QFcoaAAc7x8wsHFTK4I210V5Uk0wZnZBoJycG0hb/ftL5LmEhQBOVHSEZImSMoKCw+PCq9pb03wZe8u4WwuOxzjGUl/l/SWpB0EVdL7hMWEv5P0ZeTMSFKLcLaUFBHDvZLmhjPE6ZIaRvTvLOl/Cio2LZF0dsSyEyS9F77vbX6o5FTS52Bm9hawFUiNGO8nkt6WtDWcqf48Ytn5CgpAbw8/l1sihswmKN93d1HrlPQrSSsV3CZkmoJCKkh6P+yyJPxcBxU1RgEfAYcrvD9S+G9K2B653l9L+jTcptcUUXFK0rmSMiVtU1AkWwXeW2jMhWxbcZ+Ni1TZtfj8Ef2DoBbnOeHz44DlwL3h61kE15e3IbiYohbBdeuPExT6OIagoMe1Yf8rObgW5TMEJdfOJPhDmwycTXB9eg2CJPU1cGHYvwVBAZGkiBg+I6hIlBK+Hh0uawpsAc4Pxzo3fN0oXP4BwRVIhxHU39wOPF/E53A2sD58XoPgmvEDQPuwrQ5BBaih4WeRTnClUJtweRbQNXx+JJAeOS5B5anvgFZh+xzgyvD5hQTFmVuHY98J/C8itnz1VEvxMx0BPE9wzf6YsO1+4PawfUTY1j3chvTwM/ob8H64rGEY74Dw5z4M2AdcHW3MRX02/ijkZ1fZAfijDD+0IInmEMyW1gGPAinhslnAyIi+PyIocJES0XYZMDN8fiWFJ9FnS4jhQWB8+LwFByfROyP6Xg/8N3x+K/BcgbGmEezGNg9/6etELHuB4pPogfBz2E1QIenmiOWDgNkF3vM4cHf4/AvgWsIq9AXGzU3O9wMvhc8jk+h/CK5Pz31PDYKZ//Hh67Im0eZhXLXCf48jfxKdANwf8b66BMWcWxDetiRimQj+GFwdbcxFfTb+OPjhu/OJ60Iza2Bmx5vZ9Zb/5maRhX6PJ/iFzNIPBY8fJ5iRFidyDCR1kjRTwT2OthHcYbK4Xe2NEc+/J/hlz41nYG4sYTy5t/poAnxrZjsi3ruuhDi/sqD+aH2CW2J0j1h2PNCpwLou54fappcQzIjXhYcQuhQy/higl/LXC80d+6GIcbcSJK2mJcRbLAtK+X0K/D9gtZl9WaBLEyI+EzPLIZjJNw2XfRmxzDj4/0JpYy7NZ+Pwa+cPVQUrp+8muJ1FYZXhi7rut2D7CwQl384zs10KbgZXquOVBXxJMBP9dcEF4fG5IyXViUikzYuJ8YdgzXZLuhVYJelCM5sarus9Mzu3iPd8BPSXVIugIPRk8hdJxsy2hNt6byHb8RczK64IdFk9CzxFcBiioK8IkiGQVz/gaILbjmQREb8kkX97Sh1zaT4bF/CZ6CHOguIW04G/SqovqYaklgqqCUFwbLOZSij0S3C7ia1hAu1IcNuLsnge6CupV3hyK1nBV4qamdk6YAFwj6Taks4iqCNaKhbU6/wr8Oew6Q3gZEm/lFQrfJwuqXU4/uWSjjCzvQTHEov6itY44AyCY4m5HgNujzgJdISkgRHLvwZOLG3sBbxEcKfOyYUsewEYKilNQfHl/wd8aGZrgTeBNpIuDk/y3Uj+OwqUFDNhezSfTbXnSbR6uILg/jwrCO7j8zLB7jMEBX6XAxslfVP424HguOZISdsJklRhv+AlCndP+xOcQNlMMDsazg//FwcDnQh2Ne8mmJVF4ymCW4n0teA2Gj2BSwlmcBsJds8PC/v+Elgr6TuCwxO/KGQ8LKhRej9wVETbK+FYL4bvX0Zwy+NcI4CJ4a7zz4mCme00sxlWyP3nzewd4C7g3wQzz5bh9mHBfZ8GEhSY3gKcRP7i1iXFHKlUn43zKk7OORcTn4k651wMPIk651wMPIk651wMEv4rTg0bNrQWLVpUdhjOuUPMwoULvzGzRiX1S/gk2qJFCxYsWFDZYTjnDjGSSrrQA/Ddeeeci4knUeeci4EnUeeci4EnUeeci4EnUeeci4EnUeeci4EnUeeci0HckmhY8my+gnvqLJd0TzF9Byi4Z0+HeMXnnHNlEc+Z6G6gu5mdCqQBvSV1LthJUj2COogfxjE256qNNz9/k54v9yR1Yio9X+7Jm5+/WdkhJbS4JVEL5IQva4WPwurw3UtQu3FXvGJzrrp48/M3GfG/EWTtyMIwsnZkMeJ/IzyRxiCux0TDSuYZwCbgbTP7sMDy9sBxZvZGPONyrrp4aNFD7Nqff36ya/8uHlr0UCVFlPjimkTNbL+ZpQHNgI6S2uYuk1QDGA/8oaRxJF0jaYGkBZs3b664gJ07xGzcsTGqdleySjk7b2bZBLfV7R3RXA9oC8yStBboDLxW2MklM3vCzDqYWYdGjUossuKcCx1b59io2l3J4nl2vpGkBuHzFOAcIDN3uZltM7OGZtbCzFoA84B+ZuYlmpwrJzel30RyzeR8bck1k7kp/aZKiijxxbMUXmOCG3fVJEjek83sDUkjgQVm9locY3GuWupzYh8gODa6ccdGjq1zLDel35TX7qKX8Deq69Chg3k9UedceZO00MxK/K66X7HknHMx8CTqnHMx8CTqnHMx8CTqnHMx8CTqnHMx8CTqnHMx8CTqnHMx8CTqnHMxiOcVS1XGJx9u5INXPyNn627qHnUYXfq35OROfu2wcy561S6JfvLhRmZOymTfngMA5GzdzcxJwSX8nkidc9GqdrvzH7z6WV4CzbVvzwE+ePWzSorIOZfIql0Szdm6O6p255wrTrVLonWPOiyqduecK061S6Jd+rckqXb+zU6qXYMu/VtWUkTOuURW7U4s5Z488rPzzrnyUO2SKASJ1JOmc648VLvdeeecK0+eRF1C2vb666zu3oOVrU9hdfcebHv99coOyVVT1XJ33iW2ba+/TtZdf8Z2BfdP3/fVV2Td9WcAjujbtzJDc9WQz0Rdwtk0/sG8BJrLdu1i0/gHKykiV515EnUJZ19WVlTtzlUkT6Iu4SQ1bhxVu3MVyZOoSzjHDLsZJSfna1NyMscMu7mSInLVmZ9Ycgkn9+TRpvEPsi8ri6TGjTlm2M1+UslVCk+iLiEd0bevJ01XJfjuvHPOxcCTqHPOxSBuSVRSsqT5kpZIWi7pnkL6XCfpY0kZkuZIOiVe8TnnXFnEcya6G+huZqcCaUBvSZ0L9HnBzNqZWRpwPzAujvE551zU4nZiycwMyAlf1gofVqDPdxEv6xRc7pxzVU1cz85LqgksBH4MPGJmHxbS57fA74HaQPcixrkGuAagefPmFRavc86VJK4nlsxsf7ir3gzoKKltIX0eMbOWwK3AnUWM84SZdTCzDo0aNarYoJ1zrhiVcnbezLKBWUDvYrq9CFwYl4Ccc66M4nl2vpGkBuHzFOAcILNAn5MiXvYBVscrPuecK4t4HhNtDEwMj4vWACab2RuSRgILzOw14AZJ5wB7gW+BIXGMzznnohbPs/NLgfaFtP854vlN8YrHOefKg1+x5JxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMfAk6pxzMYhbEpWULGm+pCWSlku6p5A+v5e0QtJSSe9IOj5e8TnnXFnEcya6G+huZqcCaUBvSZ0L9FkMdDCzVOBl4P44xuecc1GLWxK1QE74slb4sAJ9ZprZ9+HLeUCzeMXnnHNlEddjopJqSsoANgFvm9mHxXS/CvhPEeNcI2mBpAWbN2+uiFCdc65U4ppEzWy/maURzDA7SmpbWD9JvwA6AGOLGOcJM+tgZh0aNWpUcQE751wJKuXsvJllA7OA3gWXSToHuAPoZ2a74xyac85FJZ5n5xtJahA+TwHOATIL9GkPPE6QQDfFKzbnnCurpDiuqzEwUVJNguQ92czekDQSWGBmrxHsvtcF/iUJ4Asz6xfHGJ1zLipxS6JmthRoX0j7nyOenxOveJxzrjz4FUvOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6LOOReDuCVRScmS5ktaImm5pHsK6fNTSYsk7ZM0IF6xOedcWcVzJrob6G5mpwJpQG9JnQv0+QK4EnghjnE551yZJcVrRWZmQE74slb4sAJ91gJIOhCvuJxzLhZxPSYqqaakDGAT8LaZfVjGca6RtEDSgs2bN5dvkM45F4W4JlEz229maUAzoKOktmUc5wkz62BmHRo1alS+QTrnXBQq5ey8mWUDs4DelbF+55wrL1EnUUl1JNUsw/saSWoQPk8BzgEyox3HOeeqkhKTqKQakgZLelPSJoLElxV+TWmspJNKua7GwExJS4GPCI6JviFppKR+4bpOl7QeGAg8Lml52TbLOefiozRn52cCM4DbgWVmdgBA0lFAN2C0pFfM7PniBjGzpUD7Qtr/HPH8I4Ljpc45lxBKk0TPMbO9kk7LTaAAZrYV+Dfwb0m1KixC55yrwkrcnTezveHTIZJejvyCvKRxBfo451y1Es2JpU1AB2CKpFWS1gDHVkxYzjmXGKK5YukXQCsz2y2pCXAfsLhiwnLOucQQzUz0S+AEADP7ysyGANdWSFTOOZcgopmJ3kRwEmkRsIjgLPqOConKOecSRKlnoma2AkgHXgTqABuB/hUUl3POJYQSZ6KSFFZgwsx2A2+Gj0L7OOdcdVKamehMSb+T1DyyUVJtSd0lTQSGVEx4zjlXtZXmmGhv4FfAPyWdAGQDKQQJeDow3swyKi5E55yrukpMoma2C3gUeDS8MqkhsDOsxOScc9VaVJXtwyuTsiooFuecSzilTqKSfl9I8zZgoe/OO+eqq2i+bN8BuA5oGj6uAc4G/iHpj+UfmnPOVX3R7M4fDaSbWQ6ApLuBl4GfAguB+8s/POecq9qimYk2B/ZEvN4LHG9mOwluh+ycc9VONDPRF4B5kl4NX/cl+NpTHWBFuUfmnHMJoNRJ1MzulfQWcBYg4DozWxAuvrwignPOuaoumrPzw4B/mdlDFRiPc84llGiOidYHpkmaLem3kn5UUUE551yiiKaK0z1m1gb4LdAEeE/SjAqLzDnnEkDU950nuE3IRmALcEz5huOcc4ml1ElU0m8kzQLeIbh+/tdmllpRgTnnXCKIZiZ6PHCzmbUxs7uBIyU9UkFxOedcQojmmOhtAJLGSFoLjAIyKygu55xLCKWpbH8ycClwGcFx0JcAmVm3Co7NOeeqvNLMRDOBHkBfMzvLzP4G7I92RZKSJc2XtETSckn3FNLnMEkvSfpU0oeSWkS7Hueci6fSfNn+EoKZ6ExJ/yW4UZ3KsK7dQHczywmLO8+R9B8zmxfR5yrgWzP7saRLgTHAoDKsy7lqZcfiTXw3bS37s3dTs8Fh1O/Vgjrt/csz8VDiTNTMXjGzQcBPgFnAMOBHkv4uqWdpV2SBnPBlrfBR8OZ2/YGJ4fOXgR6SypKwnas2dizeRPaU1ezPDuoA7c/eTfaU1exYvKmSI6seojmxtMPMJpnZBQT3nM8AbotmZZJqSsog+K7p22b2YYEuTYEvw/XtIyj6fHQh41wjaYGkBZs3b44mBOcOOd9NW4vtPZCvzfYe4LtpaysnoGqmLF+2x8y2mtnjZtY9yvftN7M0giTcUVLbAl0Km3UedCtmM3vCzDqYWYdGjRpFE4Jzh5zcGWhp2135KlMSjVV4k7tZBHcSjbQeOA5AUhJwBLA1rsE5l2BqNjgsqnZXvuKWRCU1ktQgfJ4CnMPB3zN9jR/uYT8AeNfMDpqJOud+UL9XC1Qr/6+yatWgfq8WlRNQNRPV3T5j1BiYKKkmQfKebGZvSBoJLDCz14AJwHOSPiWYgV4ax/icS0i5Z+H97HzlUKJP9Dp06GALFiwouaNzzkVB0kIz61BSv3jORONm7969rF+/nl27dlV2KC5KycnJNGvWjFq1alV2KM6VyiGZRNevX0+9evVo0aIF/jXTxGFmbNmyhfXr13PCCSdUdjjOlUqlnJ2vaLt27eLoo4/2BJpgJHH00Uf7HoRLKIdkEgU8gSYo/7lVQ0snw/i2MKJB8O/SyZUdUVQO2SRanWRnZ/Poo4/ma+vduzcNGjTgggsuyNd+5ZVXcsIJJ5CWlkZaWhoZGRkATJo0idTUVFJTUznjjDNYsmRJ3OJ31djSyfD6jbDtS8CCf1+/MaESqSfRODAzDhw4UHLHMiosiQ4fPpznnnuu0P5jx44lIyODjIwM0tLSADjhhBN47733WLp0KXfddRfXXHNNhcXrXJ53RsLenfnb9u4M2hOEJ1Fg6uINnDn6XU647U3OHP0uUxdviHnMtWvX0rp1a66//nrS09N57rnn6NKlC+np6QwcOJCcnKAWy2233cYpp5xCamoqt9xyCxDMFm+88UbOOOMMTjzxRF5++eW8cceOHcvpp59Oamoqd999d94Yn332GWlpaQwfPhyAHj16UK9evVLHe8YZZ3DkkUcC0LlzZ9avXx/zZ+BcibYV8f+sqPYqqNon0amLN3D7lI/ZkL0TAzZk7+T2KR+XSyJdtWoVV1xxBW+//TYTJkxgxowZLFq0iA4dOjBu3Di2bt3KK6+8wvLly1m6dCl33nln3nuzsrKYM2cOb7zxBrfdFtR5mT59OqtXr2b+/PlkZGSwcOFC3n//fUaPHk3Lli3JyMhg7NixJcZ1xx13kJqayrBhw9i9++DrqydMmMB5550X8/Y7V6IjmkXXXgVV+yQ6dtoqdu7NX2N65979jJ22Kuaxjz/+eDp37sy8efNYsWIFZ555JmlpaUycOJF169ZRv359kpOTufrqq5kyZQqHH3543nsvvPBCatSowSmnnMLXX38NBEl0+vTptG/fnvT0dDIzM1m9enVUMd13331kZmby0UcfsXXrVsaMGZNv+cyZM5kwYcJB7c5ViB5/hlop+dtqpQTtCeKQ/J5oNL7K3hlVezTq1KkDBMdEzz33XP75z38e1Gf+/Pm88847vPjiizz88MO8++67ABx22A/FI3KvKjMzbr/9dq699tp8Y6xdu7bUMTVu3Dhv/KFDh/LAAw/kLVu6dClXX301//nPfzj66IMqEDpX/lJ/Hvz7zshgF/6IZkECzW1PANV+JtqkQUpU7WXRuXNn5s6dy6effgrA999/zyeffEJOTg7btm3j/PPP58EHH8w7U16UXr168dRTT+UdT92wYQObNm2iXr16bN++vVSxZGVlAUFCnjp1Km3bBtUIv/jiCy6++GKee+45Tj755LJuqnPRS/05DFsGI7KDfxMogYLPRBneqxW3T/k43y59Sq2aDO/VqtzW0ahRI5555hkuu+yyvGOQo0aNol69evTv359du3ZhZowfP77YcXr27MnKlSvp0qULAHXr1uX555+nZcuWnHnmmbRt25bzzjuPsWPH0rVrVzIzM8nJyaFZs2ZMmDCBXr16cfnll7N582bMjLS0NB577DEARo4cyZYtW7j++usBSEpKwmsSOFeyQ7IAycqVK2ndunWpx5i6eANjp63iq+ydNGmQwvBerbiwfdPyDtWVUrQ/P+cqQrUuQBKtC9s39aTpnCuTan9M1DnnYuFJ1DnnYuBJ1DnnYuBJ1DnnYuBJ1DnnYuBJ9BAQTSm8NWvW0KlTJ0466SQGDRrEnj17AHj//fdJT08nKSkpX8ET51zxPInGQVUqhXfrrbcybNgwVq9ezZFHHsmECRMAaN68Oc888wyDBw+usDidOxR5EoUKqaxdFUvhmRnvvvsuAwYMAGDIkCFMnToVgBYtWpCamkqNGv5fwrlo+Jftcytr5xaGza2sDTFfw7tq1SqefvppRo4cycUXX8yMGTOoU6cOY8aMYdy4cdxwww288sorZGZmIons7Oy89+aWwsvMzKRfv34MGDAgXyk8M6Nfv355pfCWLVtW4rX3W7ZsoUGDBiQlBT/2Zs2asWFD7CX/nKvOPIkWV1k7xiSaWwrvjTfeyCuFB7Bnzx66dOmSrxRenz598h2/LKkUHkBOTg6rV6+mefPmpYqnsEt8/Z5GzsXGk2gFVtauaqXwGjZsSHZ2Nvv27SMpKYn169fTpEmTsmyacy7kB8DiUFm7qpTCk0S3bt3yjrFOnDiR/v37x7h1zlVvcUuiko6TNFPSSknLJd1USJ8jJb0iaamk+ZLaVnhgcaisHVkKLzU1lc6dO5OZmcn27du54IILSE1N5Wc/+1mpSuENHjyYLl260K5dOwYMGMD27ds5+uij80rh5Z5Y6tq1KwMHDuSdd96hWbNmTJs2DSDveOyPf/xjtmzZwlVXXQXARx99RLNmzfjXv/7FtddeS5s2bcpt+507lMWtFJ6kxkBjM1skqR6wELjQzFZE9BkL5JjZPZJ+AjxiZj2KG7c8SuGxdHJCV9Y+1HgpPFcVVLlSeGaWBWSFz7dLWgk0BVZEdDsFuC/skymphaQfmdnXFRpc6s89aTrnyqRSjolKagG0Bz4ssGgJcHHYpyNwPHDQwUlJ10haIGnB5s2bKzZY55wrRtyTqKS6wL+Bm83suwKLRwNHSsoAfgcsBvYVHMPMnjCzDmbWoVGjRhUes3POFSWuX3GSVIsggU4ysykFl4dJdWjYV8Ca8OGcc1VSPM/OC5gArDR+pGfKAAAa1UlEQVSzcUX0aSCpdvjyauD9QmarzjlXZcRzJnom8Evg43B3HeBPQHMAM3sMaA08K2k/wQmnq+IYn3PORS2eZ+fnAMVeY2hmHwAnxSeiQ0d2djYvvPBC3u2OAWrWrEm7du2AoELTa6+9BgSl8C699FK2bt2aVxildu3ahY7rnCuZX7EUB5VRCi8lJYWMjAwyMjLyEigUXQrPOVc2nkSBNz9/k54v9yR1Yio9X+7Jm5+/GfOYlV0KrzDFlcJzzpVNtS9A8ubnbzLifyPYtX8XAFk7shjxvxEA9DmxT0xjV2YpvF27dtGhQweSkpK47bbbuPDCC70UnnMVoNon0YcWPZSXQHPt2r+LhxY9FHMSrcxSeF988QVNmjTh888/p3v37rRr14769esf1M9L4TkXm2qfRDfu2BhVezQqsxRebom7E088kbPPPpvFixdzySWXeCk858pZtT8memydY6NqL4t4l8L79ttv2b17NwDffPMNc+fO5ZRTTvFSeM5VgGo/E70p/aZ8x0QBkmsmc1P6QZX6yiyyFF5uchs1ahT16tWjf//+7Nq1CzMrVSm8lStX0qVLFwDq1q3L888/T8uWLfNK4Z133nlcdNFFXHvttdSoUYMDBw7knbyCoBTepZdeyp133kn79u3zSuE558ombqXwKkp5lMJ78/M3eWjRQ2zcsZFj6xzLTek3xXw81JWdl8JzVUGVK4VXlfU5sY8nTedcmVT7Y6LOORcLT6LOORcDT6LOORcDT6LOORcDT6LOORcDT6KHgIJVnNatW8dpp51GWloabdq04bHHHjvoPf369aNt24q/I7VzhzpPonEQ71J4jRs35n//+x8ZGRl8+OGHjB49mq+++ipv+ZQpU6hbt26FxeNcdeJJFNj2+uus7t6Dla1PYXX3Hmx7/fWYx6zMUni1a9fOu/Z+9+7d+RJ4Tk4O48aN484774x5G51zBLOkRH6cdtppVtCKFSsOaitK9muv2cpT02xFq5/kPVaemmbZr71W6jEKs2bNGpNkH3zwgW3evNm6du1qOTk5ZmY2evRou+eee2zLli128skn24EDB8zM7NtvvzUzsyFDhtiAAQNs//79tnz5cmvZsqWZmU2bNs1+/etf24EDB2z//v3Wp08fe++992zNmjXWpk2bfOv/4osvrF27dpaSkmIPP/xwXvvNN99sU6ZMKfQ9VUU0Pz/nKgqwwEqRg6r9THTT+AexXflL4dmuXWwa/2DMY+eWwps3b15eKby0tDQmTpzIunXr8pXCmzJlCocffnjee0sqhZeenk5mZiarV68udN3HHXccS5cu5dNPP2XixIl8/fXXZGRk8Omnn3LRRRfFvG3OuUC1v+xzX1ZWVO3RqMxSeLmaNGlCmzZtmD17Nps3b2bhwoW0aNGCffv2sWnTJs4++2xmzZoV45Y6V31V+5loUuPGUbWXRbxL4a1fv56dO3cCQVm8uXPn0qpVK37zm9/w1VdfsXbtWubMmcPJJ5/sCdS5GFX7megxw24m664/59ulV3Iyxwy7udzWEe9SeD179uQPf/gDkjAzbrnllrw7fzrnypeXwiM4O79p/IPsy8oiqXFjjhl2M0f07VveobpS8lJ4rirwUnhROKJvX0+azrkyqfbHRJ1zLhaeRJ1zLgZxS6KSjpM0U9JKScslHXQTI0lHSHpd0pKwz9B4xeecc2URz2Oi+4A/mNkiSfWAhZLeNrMVEX1+C6wws76SGgGrJE0ysz1xjNM550otbjNRM8sys0Xh8+3ASqBpwW5APUkC6gJbCZKvc85VSZVyTFRSC6A98GGBRQ8DrYGvgI+Bm8ys4sofHSIKVnEC6N27Nw0aNOCCCy7I1/7OO++Qnp5OWloaZ511Vt4FAM65sol7EpVUF/g3cLOZfVdgcS8gA2gCpAEPS6pfyBjXSFogacHmzZsrPOZYWZxL4QEMHz6c55577qC+v/nNb5g0aRIZGRkMHjyYUaNGVVhczlUHcU2ikmoRJNBJZjalkC5DgSlhEZVPgTXATwp2MrMnzKyDmXVo1KhRzHF98uFGJv5pLo9c9y4T/zSXTz7cGPOYlVkKD6BHjx7Uq1fvoLgk8d13wd+ubdu20aRJk5i31bnqLG4nlsLjnBOAlWY2rohuXwA9gNmSfgS0Aj6vyLg++XAjMydlsm9PMFPM2bqbmZMyATi507Exjb1q1SqefvppRo4cycUXX8yMGTOoU6cOY8aMYdy4cdxwww288sorZGZmIons7Oy892ZlZTFnzhwyMzPp168fAwYMYPr06axevZr58+djZvTr14/333+f0aNHs2zZshKvvQd48sknOf/880lJSaF+/frMmzcvpm10rrqL50z0TOCXQHdJGeHjfEnXSbou7HMvcIakj4F3gFvN7JuKDOqDVz/LS6C59u05wAevfhbz2JVZCq8o48eP56233mL9+vUMHTqU3//+9zFvp3PVWdxmomY2B1AJfb4CesYnokDO1t1RtUejKpTCi7R582aWLFlCp06dABg0aBC9e/eOeruccz+o9lcs1T3qsKjayyLepfCKcuSRR7Jt2zY++eQTAN5++20v9OFcjKp9AZIu/VvmOyYKkFS7Bl36tyy3dcS7FN7YsWPp2rUrmZmZ5OTk0KxZMyZMmECvXr34xz/+wSWXXEKNGjU48sgjeeqpp8ptO52rjrwUHsHJpQ9e/Yycrbupe9RhdOnfMuaTSq7svBSeqwq8FF4UTu50rCdN51yZVPtjos45FwtPos45FwNPos45FwNPos45FwNPos45FwNPohXk//7v/2jdujWXX345DzzwQGWHw9q1a3nhhRfyXm/ZsoVu3bpRt25dbrjhhnx9zz77bFq1akVaWhppaWls2rQJgHHjxuUVS+nRowfr1q2L6zY4VxV5Eq0gjz76KG+99RYnnXRSiX33799f4fEUTKLJycnce++9RSb43HJ5GRkZHHPMMQC0b9+eBQsWsHTpUgYMGMAf//jHCo/buarOkyiwcvZMnvjtUP56aV+e+O1QVs6eGdN41113HZ9//jn9+vVj/PjxLFmyhO7du3PSSSfxj3/8A4BZs2bRrVs3Bg8eTLt27QB4/vnn6dixI2lpaVx77bXs37+f/fv3c+WVV9K2bVvatWuXd1XT2Wefza233krHjh05+eSTmT17NhAk5OHDh+eVy3v88ceBoFze7NmzSUtLY/z48dSpU4ezzjqL5OTkUm9Xt27d8oqkdO7cmfXr18f0OTl3KKj2X7ZfOXsm0594mH17gssxt3+zmelPPAxA667dyjTmY489xn//+19mzpzJww8/zCuvvMK8efPYsWMH7du3p0+fPkBQfGTZsmWccMIJrFy5kpdeeom5c+dSq1Ytrr/+eiZNmkSbNm3YsGEDy5YtA8hXLm/fvn3Mnz+ft956i3vuuYcZM2YwYcIEjjjiCD766CN2797NmWeeSc+ePRk9ejQPPPAAb7zxRqm2YejQodSsWZNLLrmEO++8k6CS4Q8mTJjAeeedV6bPx7lDSbVPorNffDYvgebat2c3s198tsxJtKD+/fuTkpJCSkoK3bp1Y/78+TRo0ICOHTtywgknAMFtOxYuXMjpp58OwM6dOznmmGPo27cvn3/+Ob/73e/o06cPPXv+UOTq4osvBuC0007Lq+Q0ffp0li5dmlfIedu2baxevZratWuXOt5JkybRtGlTtm/fziWXXMJzzz3HFVdckbf8+eefZ8GCBbz33nsxfS7OHQqqfRLdvqXwcqVFtZdFwVlc7uvcUnkQlLkbMmQI991330HvX7JkCdOmTeORRx5h8uTJeUVDcsvl1axZk3379uWN87e//Y1evXrlG2PWrFmljrdp0+D+gfXq1WPw4MHMnz8/L4nOmDGDv/zlL7z33nv5yvU5V11V+2Oi9Y5uGFV7Wbz66qvs2rWLLVu2MGvWrLzZZqQePXrw8ssv550J37p1K+vWreObb77hwIEDXHLJJdx7770sWrSo2HX16tWLv//97+zduxeATz75hB07dpS6XN6+ffv45pvgD8jevXt54403aNu2LQCLFy/m2muv5bXXXss72eRcdVftZ6JdL70i3zFRgKTah9H10iuKeVd0OnbsSJ8+ffjiiy+46667aNKkSV5Nz1ynnHIKo0aNomfPnhw4cIBatWrxyCOPkJKSwtChQ/NudFfYTDXS1Vdfzdq1a0lPT8fMaNSoEVOnTiU1NZWkpCROPfVUrrzySoYNG0aLFi347rvv2LNnD1OnTmX69Okcf/zx9OrVi71797J//37OOeccfv3rXwPBze9ycnIYOHAgAM2bN+e1114rt8/JuUTkpfAITi7NfvFZtm/5hnpHN6TrpVeU2/FQFz0vheeqAi+FF4XWXbt50nTuEDR18QbGTlvFV9k7adIgheG9WnFh+6blug5Pos65Q9LUxRu4fcrH7NwbXMyyIXsnt0/5GKBcE2m1P7HknDs0jZ22Ki+B5tq5dz9jp60q1/V4EnXOHZK+yt4ZVXtZeRJ1zh2SmjRIiaq9rDyJOucOScN7tSKlVs18bSm1ajK8V6tyXY8n0Spg9+7dnHPOOaSlpfHSSy8xe/Zs2rRpQ1paGjt3lu+uR1llZ2fz6KOP5mvr3bs3DRo04IILLsjXvmbNGjp16sRJJ53EoEGD2LNnDwDvv/8+6enpJCUl5V2W6lxFubB9U+67uB1NG6QgoGmDFO67uF25n533JFoFLF68mL1795KRkcGgQYOYNGkSt9xyCxkZGaSklLzrYWZ5X8avKIUl0eHDh/Pcc88d1PfWW29l2LBhrF69miOPPJIJEyYAwZfzn3nmGQYPHlyhsbqqrbyrphXnwvZNmXtbd9aM7sPc27qXewKFOCZRScdJmilppaTlkm4qpM9wSRnhY5mk/ZKOqujYdizeRNbo+ay/bTZZo+ezY/Gmchn32WefJTU1lVNPPZVf/vKXvP7663Tq1In27dtzzjnn8PXXX7Np0yZ+8YtfkJGRQVpaGo8//jiTJ09m5MiRXH755QCMHTs2r7Td3XffDQT1QVu3bs31119Peno6X375JdOnT6dLly6kp6czcOBAcnJygKAMXm4x5VtuuQWAK6+8khtvvJEzzjiDE088Md/MsLD13XbbbXz22WekpaUxfPhwILhUtV69evm22cx49913GTBgAABDhgxh6tSpALRo0YLU1FRq1PC/3dVVbtW07d9sBrO8qmkVmUgrWjy/J7oP+IOZLZJUD1go6W0zW5HbwczGAmMBJPUFhpnZ1ooMasfiTWRPWY3tDWZy+7N3kz1lNQB12pf9+vDly5fzl7/8hblz59KwYUO2bt2KJObNm4cknnzySe6//37++te/8uSTT+YrU/fBBx9wwQUXMGDAAKZPn87q1auZP38+Zka/fv14//33ad68OatWreLpp5/m0Ucf5ZtvvmHUqFHMmDGDOnXqMGbMGMaNG8cNN9zAK6+8QmZmJpLyldLLyspizpw5ZGZm0q9fv2LXN3r0aJYtW0ZGRkax271lyxYaNGhAUlLwX6tZs2Zs2LChzJ+jO7TEo2pavMUtiZpZFpAVPt8uaSXQFFhRxFsuA/5Z0XF9N21tXgLNZXsP8N20tTEl0dzZWMOGQSGTo446io8//phBgwaRlZXFnj178srgFWf69OlMnz6d9u3bA5CTk8Pq1atp3rw5xx9/PJ07dwZg3rx5rFixgjPPPBOAPXv20KVLF+rXr09ycjJXX301ffr0yXf88sILL6RGjRqccsopfP311yWurzQKu4y4YBUrV33Fo2pavFXKFUuSWgDtgQ+LWH440Bu4obDl5Wl/9u6o2kvLzA5KHr/73e/4/e9/T79+/Zg1axYjRowo1Ti333471157bb72tWvXHlRK79xzz+Wf/zz47878+fN55513ePHFF3n44Yd59913AfKVsstNfsWtrzQaNmxIdnY2+/btIykpifXr19OkSZNSvdcd+uod3TDYlS+kPVHF/eCUpLrAv4Gbzey7Irr1BeYWtSsv6RpJCyQt2Lz54B9INGo2KLwmZlHtpdWjRw8mT57Mli1bgKC03bZt2/JqdU6cOLFU4/Tq1Yunnnoq7/jmhg0b8srlRercuTNz587l008/BeD777/nk08+IScnh23btnH++efz4IMPlrg7XtT6SltKTxLdunXLO8Y6ceJE+vfvX6ptdYe+rpdeQVLt/L9b5V01Ld7iOhOVVIsggU4ysynFdL2UYnblzewJ4AkIqjjFElP9Xi3yHRMFUK0a1O/VIpZhadOmDXfccQc/+9nPqFmzJu3bt2fEiBEMHDiQpk2b0rlzZ9asWVPiOD179mTlypV06dIFgLp16/L8889Ts2b+7781atSIZ555hssuu4zdu4NZ9KhRo6hXrx79+/dn165dmFnePZqiXV/Lli0588wzadu2Leeddx5jx46la9euZGZmkpOTQ7NmzZgwYQK9evVizJgxXHrppdx55520b9+eq666CoCPPvqIiy66iG+//ZbXX3+du+++m+XLl0f92brElXvc81Cqmha3UngK9m0nAlvN7OZi+h0BrAGOM7MdJY1bHqXwdizexHfT1rI/ezc1GxxG/V4tYjoe6mLjpfBcVVAVS+GdCfwS+FhS7j7ln4DmAGb2WNh2ETC9NAm0vNRpf4wnTedcmcTz7PwcoMTTtGb2DPBMRcfjnHPlwb/17JxzMThkk2ii3/akuvKfm0s0h2QSTU5OZsuWLf4LmWDMjC1btpCcnFzZoThXaofk7UGaNWvG+vXrifU7pC7+kpOTadasWWWH4VypHZJJtFatWqW6pNI552J1SO7OO+dcvHgSdc65GHgSdc65GMTtss+KImk7UL73QK1cDYHErQt2MN+eqs23p2jHm1mjkjodCieWVpXm+tZEIWmBb0/V5dtTtVXG9vjuvHPOxcCTqHPOxeBQSKJPVHYA5cy3p2rz7ana4r49CX9iyTnnKtOhMBN1zrlKk3BJVNKw8L71yyT9U1KypO6SFoVtEyUlzLcOJN0Uxr1c0s1h21GS3pa0Ovz3yMqOs7SK2J6B4esDkhLqTHAR2zNWUqakpZJekdSgsuMsrSK2595wWzIkTZeUMHcWLGx7IpbdIskkVexd8MwsYR4Et1heA6SErycDvwK+BE4O20YCV1V2rKXcnrbAMuBwgq+bzQBOAu4Hbgv73AaMqexYY9ye1kArYBbQobLjLIft6QkkhX3GHAI/n/oRfW4EHqvsWGPZnnDZccA0YB3QsCLjSLiZKMGHlRLONg8HdgC7zeyTcPnbwCWVFVyUWgPzzOx7M9sHvEdwe5T+BPejIvz3wkqKL1qFbo+ZrTSzRLwgoqjtmR6+BpgHJErZqaK2J/Kuu3WARDlRUtTvD8B44I/EYVsSKoma2QbgAeALIAvYRjAbrRWxmziA4K9QIlgG/FTS0ZIOB84niP1HZpYFEP6bKDeAKmp7ElVptudXwH/iHlnZFLk9kv4i6UvgcuDPlRhjNArdHkn9gA1mtiQeQSTMsUOA8Nhgf+AEIBv4F8EP/VJgvKTDgOnAviIHqULMbKWkMQSz5xxgCQkSe2Gq2/ZIuiN8PalyIoxOcdtjZncAd0i6HbgBuLvSAi2lYrbnDoJDLnGRUDNR4BxgjZltNrO9wBTgDDP7wMy6mllH4H1gdaVGGQUzm2Bm6Wb2U2ArQexfS2oMEP67qTJjjEYR25OwitoeSUOAC4DLLTwIlwhK8fN5gcQ5HFbY9qwlmGQtkbSW4FDLIknHVlQMiZZEvwA6Szo8vI99D2ClpGMAwpnorcBjxYxRpUTE3hy4GPgn8BowJOwyBHi1cqKLXhHbk7AK2x5JvQn+n/Uzs+8rM75oFbE9J0V06QdkVkZsZVHI9jxrZseYWQszawGsB9LNbGNFxZBQu/Nm9qGkl4FFBNP2xQRXKIySdAHBH4W/m9m7lRhmtP4t6WhgL/BbM/tW0mhgsqSrCP5wDKzUCKNT2PZcBPwNaAS8KSnDzHpVapSlV9j2PAwcBrwd/C1nnpldV5lBRqGw7XlSUivgAMHZ7ETZFihke+IdgF+x5JxzMUi03XnnnKtSPIk651wMPIk651wMPIk651wMPIk651wMPIm6ChVekpcRPjZK2hDxunYU4/yquC9MS6otaauke8sncudKx7/i5OJG0gggx8weKMN75wA3mFlGEcv7ERScOMbMTo4p0OLjSIooPuKcz0Rd5ZE0RNL8cFb6qKQakpIkPSfp47BO5I2SBgFpwEvFzGAvA8YRXDJ7esQ6Okn6QNISSR+GV7slSRofjr9U0vVh3/W5tUEldZY0I3w+StLjkt4GnpbUUtJsSYslLZTUKWJ9fwpjXxIW9WglaX7E8taRr13iS6grltyhQ1JbgrJlZ5jZPklPEBSS+Yyg/mO7sF8DM8uW9DuKmIlKqgP8DBgKHEuQUD+SlAy8CFxiZoskHQHsBq4HmgCnmtl+SUeVIuT2wE/NbFdYMejc8PlPCMoVdpLUFzgP6GhmOyUdZWZbJe2S1NbMloUxPl3Wz81VPT4TdZXlHOB0YIGkDIIk2BL4FGgl6SFJvQjKHZakH/C2me0iqOx1iaQaBPUmvzCzRQBmts3M9ofrfix8jpltLcU6Xg3Hh+CSzwmSlhEk6VMitukpM9tZYNwJwFAFNXAHkuD1BFx+PhN1lUUECeeugxZIqQQzuhsJKgpdU8JYlxHMBNeGr48Bfgp8R+FFeVVE+z5+mFgkF1i2I+L5HwjupvALoBZBGbbixv0X8CdgLvCBmWUXsy0uwfhM1FWWGcDPFd7/JjyL31xSI4ITnv8iqGmZHvbfDtQrOIiCGrOdgGYRlXtuJEisy4HjJaWHfetLqklQc/Y34XMidufXAqeFz4srB3cEkBWWwBtCkDwJx71KUkrkuGGlp3eBh/Fd+UOOJ1FXKczsY+AeYIakpQQJ6EcEldbfD3fx/0Ewg4Mg+TxZyImlSwh25fdGtE0lON56gCCZ/l3SknAdhwGPAxuBpWH7z8P3jQAelTQb2FNM+A8DV0uaBxxPcJwVM3sD+C8/HKIYFvGeSQSVht4pxcfjEoh/xcm5OJB0G3CYmd1T2bG48uXHRJ2rYJJeJ5hhd6/sWFz585moc87FwI+JOudcDDyJOudcDDyJOudcDDyJOudcDDyJOudcDDyJOudcDP4/glPsdZbBQN8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x, y = [], []\n",
    "for name in df.modelname:\n",
    "    if 'resnet' in name and 'inception' not in name:\n",
    "        #print(name)\n",
    "        mf = df[df.modelname==name]\n",
    "        x = mf['acc5'].values\n",
    "        y = mf['avg_alpha'].values\n",
    "        if y > 0:\n",
    "            plt.scatter(x,y, label=name)\n",
    "        \n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha])$\"+\"\\nPretrained ResNet  Models\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 288,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1a670b8860>"
      ]
     },
     "execution_count": 288,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x, y = [], []\n",
    "for name in df.modelname: \n",
    "    if 'inception' in name and 'resnet' not in name:\n",
    "        #print(name)\n",
    "        mf = df[df.modelname==name]\n",
    "        x = mf['acc5'].values\n",
    "        y = mf['avg_alpha'].values\n",
    "        if y > 0:\n",
    "            plt.scatter(x,y, label=name)\n",
    "        \n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha])$\"+\"\\nPretrained Inception3,4 Models\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#for mf inacc = [df_vgg11, df_vgg13]:\n",
    "    x = mf['acc5'].values\n",
    "    y = mf['avg_alpha'].values\n",
    "    label = mf['legend'].values[0]\n",
    "    print(label)\n",
    "    plt.scatter(x,y,label=label)\n",
    "\n",
    "plt.legend()\n",
    "plt.title(r\"Accuracy vs Avg$(\\alpha)$\"+\"\\nPretrained ImageNet Models\")\n",
    "plt.xlabel(r\"Test Accuracy\")\n",
    "plt.ylabel(r\"Avg$(\\alpha)$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot for each layer .. BIG SCATTER PLOT"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### TODO\n",
    "\n",
    "- CHECK for RANK COLLAPSE\n",
    "\n",
    "- Find more pretrained variants ?  NLP models...\n",
    "\n",
    "- Test transfer learning power laws\n",
    "\n",
    "- Yolov3 versions \n",
    "https://medium.com/@jonathan_hui/real-time-object-detection-with-yolo-yolov2-28b1b93e2088"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  },
  "widgets": {
   "state": {
    "52ecdae62c394a67868cd3ee83e19861": {
     "views": [
      {
       "cell_index": 14
      }
     ]
    },
    "e232317699e24554b5ed09fa6c78c662": {
     "views": [
      {
       "cell_index": 8
      }
     ]
    }
   },
   "version": "1.2.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}