Math Archive Prediction Tasks with Transformers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "0f329f52",
   "metadata": {},
   "source": [
    "So far, based on the data of math-related arXiv submissions—which we scraped and preprocessed in `Scarping and Cleaning the Data.ipynb`—we considered the text classification tasks of predicting the MSC classes and the primary arXiv category based on `cleaned_text` which is obtained from putting together the preprocessed title and abstract. (These are multi-label and multi-class tasks respectively). Support vector machines and convolutional neural networks were used in notebooks `MSC Prediction.ipynb` and `Math Archive Primary Category Prediction.ipynb` for these tasks respectively. Here, we play with the more advanced *transformer architecture* for the same classification problems, data, and train/test splits. We shall [implement transformers in the Keras API](https://keras.io/examples/nlp/text_classification_with_transformer). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "9d8b763b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import json\n",
    "import gc\n",
    "\n",
    "import keras\n",
    "from keras import backend as K     # For Keras backend utilities.\n",
    "from keras import ops              # For numpy operations. \n",
    "from keras import layers           \n",
    "\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "\n",
    "from sklearn.metrics import f1_score, accuracy_score, jaccard_score"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "97b22c5d",
   "metadata": {},
   "source": [
    "We begin with classifying the primary arXiv category with transformers. Previously, in `Math Archive Primary Category Prediction.ipynb` we trained a convolutional neural net for this:\n",
    "\n",
    "$\\hspace{3cm}$  `cleaned_text`  $\\mapsto$ the primary arXiv category (e.g. `math.AG`) $\\hspace{1cm}$  (~30 labels).\n",
    "\n",
    "We first load the data and the train/test split saved from before. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "52a6ff29",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training instances: 119381\n",
      "Number of test instances: 39794\n",
      "Number of the labels: 31\n"
     ]
    }
   ],
   "source": [
    "data=pd.read_csv('data/data_encoded_CatPred.zip',compression='zip')\n",
    "\n",
    "with open('data/train_index_CatPred.json','r') as file:\n",
    "    train_index=json.load(file)\n",
    "X_train=data.iloc[train_index]['cleaned_text']\n",
    "y_train=data.iloc[train_index,1:]\n",
    "\n",
    "with open('data/test_index_CatPred.json','r') as file:\n",
    "    test_index=json.load(file)\n",
    "X_test=data.iloc[test_index]['cleaned_text']\n",
    "y_test=data.iloc[test_index,1:]\n",
    "\n",
    "print(f'Number of training instances: {y_train.shape[0]}')\n",
    "print(f'Number of test instances: {y_test.shape[0]}')\n",
    "print(f'Number of the labels: {y_train.shape[1]}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f0637a7",
   "metadata": {},
   "source": [
    "Implementing the positional embedding and the transformer block (including [Multi-Head Attention](https://keras.io/api/layers/attention_layers/multi_head_attention/), MLP, Add & Norm components):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "4ebb571c",
   "metadata": {},
   "outputs": [],
   "source": [
    "class TokenAndPositionEmbedding(layers.Layer):\n",
    "    def __init__(self, maxlen, input_dim, embed_dim):\n",
    "        super().__init__()\n",
    "        self.token_emb = layers.Embedding(input_dim=input_dim, output_dim=embed_dim)\n",
    "        self.pos_emb = layers.Embedding(input_dim=maxlen, output_dim=embed_dim)\n",
    "\n",
    "    def call(self, x):\n",
    "        maxlen = ops.shape(x)[-1]\n",
    "        positions = ops.arange(start=0, stop=maxlen, step=1)\n",
    "        positions = self.pos_emb(positions)\n",
    "        x = self.token_emb(x)\n",
    "        return x + positions\n",
    "\n",
    "class TransformerBlock(layers.Layer):\n",
    "    def __init__(self, embed_dim, num_heads, ff_dim, rate=0.1):\n",
    "        super().__init__()\n",
    "        self.att = layers.MultiHeadAttention(num_heads=num_heads, key_dim=embed_dim)\n",
    "        self.ffn = keras.Sequential(\n",
    "            [layers.Dense(ff_dim, activation=\"relu\"), layers.Dense(embed_dim),]\n",
    "        )\n",
    "        self.layernorm1 = layers.LayerNormalization(epsilon=1e-6)\n",
    "        self.layernorm2 = layers.LayerNormalization(epsilon=1e-6)\n",
    "        self.dropout1 = layers.Dropout(rate)\n",
    "        self.dropout2 = layers.Dropout(rate)\n",
    "\n",
    "    def call(self, inputs):\n",
    "        attn_output = self.att(inputs, inputs)\n",
    "        attn_output = self.dropout1(attn_output)\n",
    "        out1 = self.layernorm1(inputs + attn_output)\n",
    "        ffn_output = self.ffn(out1)\n",
    "        ffn_output = self.dropout2(ffn_output)\n",
    "        return self.layernorm2(out1 + ffn_output)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "51aded1e",
   "metadata": {},
   "source": [
    "We next tokenize and then pad the data with the same parameters used before for this text data (`maxlen=100` , `padding='post'`)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "729f0373",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary size (based on the training set): 71614\n"
     ]
    }
   ],
   "source": [
    "maxlen=100\n",
    "\n",
    "tokenizer=Tokenizer()\n",
    "tokenizer.fit_on_texts(X_train)\n",
    "\n",
    "X_train=tokenizer.texts_to_sequences(X_train)\n",
    "X_train=pad_sequences(X_train, maxlen=maxlen, padding='post')\n",
    "\n",
    "X_test=tokenizer.texts_to_sequences(X_test)\n",
    "X_test=pad_sequences(X_test, maxlen=maxlen, padding='post')\n",
    "\n",
    "# Saving the number of tokens \n",
    "vocabulary_size=len(tokenizer.word_counts)\n",
    "print(f'Vocabulary size (based on the training set): {vocabulary_size}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "589b0cbf",
   "metadata": {},
   "source": [
    "Below, we set up the transformer model. (Keras' Functional API is used instead of the Sequential API, [a comparison](https://medium.com/analytics-vidhya/keras-model-sequential-api-vs-functional-api-fc1439a6fb10).) Instead of a convolutional later, we include an attention block with two heads. Other architecture parameters are more or less the same as what used before for training a convolutional neural network for the same problem: the embedding layer outputs vectors of dimension 50; and a ReLU layer of size 200 is present. Dropout layers are used too (a higher drop out rate for more complex parts such as the attention block and the ReLU layer). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "55084c9c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\khash\\anaconda3\\Lib\\site-packages\\keras\\src\\backend\\tensorflow\\core.py:192: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional_1\"</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1mModel: \"functional_1\"\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ input_layer (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ token_and_position_embedding    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">50</span>)        │     <span style=\"color: #00af00; text-decoration-color: #00af00\">3,585,750</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">TokenAndPositionEmbedding</span>)     │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">50</span>)        │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ transformer_block               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">50</span>)        │        <span style=\"color: #00af00; text-decoration-color: #00af00\">27,064</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">TransformerBlock</span>)              │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling1d        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">50</span>)             │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">GlobalAveragePooling1D</span>)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_4 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">50</span>)             │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_2 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">200</span>)            │        <span style=\"color: #00af00; text-decoration-color: #00af00\">10,200</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_5 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">200</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">31</span>)             │         <span style=\"color: #00af00; text-decoration-color: #00af00\">6,231</span> │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                   \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ input_layer (\u001b[38;5;33mInputLayer\u001b[0m)        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m)            │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ token_and_position_embedding    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m50\u001b[0m)        │     \u001b[38;5;34m3,585,750\u001b[0m │\n",
       "│ (\u001b[38;5;33mTokenAndPositionEmbedding\u001b[0m)     │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (\u001b[38;5;33mDropout\u001b[0m)               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m50\u001b[0m)        │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ transformer_block               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m50\u001b[0m)        │        \u001b[38;5;34m27,064\u001b[0m │\n",
       "│ (\u001b[38;5;33mTransformerBlock\u001b[0m)              │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling1d        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m50\u001b[0m)             │             \u001b[38;5;34m0\u001b[0m │\n",
       "│ (\u001b[38;5;33mGlobalAveragePooling1D\u001b[0m)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_4 (\u001b[38;5;33mDropout\u001b[0m)             │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m50\u001b[0m)             │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_2 (\u001b[38;5;33mDense\u001b[0m)                 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m200\u001b[0m)            │        \u001b[38;5;34m10,200\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_5 (\u001b[38;5;33mDropout\u001b[0m)             │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m200\u001b[0m)            │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_3 (\u001b[38;5;33mDense\u001b[0m)                 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m31\u001b[0m)             │         \u001b[38;5;34m6,231\u001b[0m │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">3,629,245</span> (13.84 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m3,629,245\u001b[0m (13.84 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">3,629,245</span> (13.84 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m3,629,245\u001b[0m (13.84 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "embed_dim = 50            # Embedding size for each token\n",
    "num_heads = 2             # Number of attention heads\n",
    "ff_dim = 64               # Hidden layer size in feed forward network inside transformer\n",
    "\n",
    "inputs = layers.Input(shape=(maxlen,))\n",
    "embedding_layer = TokenAndPositionEmbedding(maxlen=maxlen, input_dim=vocabulary_size+1, embed_dim=embed_dim)\n",
    "x = embedding_layer(inputs)\n",
    "x = layers.Dropout(0.1)(x)                                           \n",
    "transformer_block = TransformerBlock(embed_dim, num_heads, ff_dim,rate=0.2)\n",
    "x = transformer_block(x)\n",
    "x = layers.GlobalAveragePooling1D()(x)\n",
    "x = layers.Dropout(0.2)(x)\n",
    "x = layers.Dense(200, activation=\"relu\")(x)\n",
    "x = layers.Dropout(0.1)(x)\n",
    "outputs = layers.Dense(31, activation=\"softmax\")(x)\n",
    "\n",
    "model = keras.Model(inputs=inputs, outputs=outputs)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "932fcac6",
   "metadata": {},
   "source": [
    "We use early stopping with patience, and we keep the best model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "af8c6352",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m153s\u001b[0m 56ms/step - accuracy: 0.1068 - loss: 3.1998 - val_accuracy: 0.3844 - val_loss: 2.1678\n",
      "Epoch 2/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m162s\u001b[0m 62ms/step - accuracy: 0.4260 - loss: 2.0159 - val_accuracy: 0.5497 - val_loss: 1.5346\n",
      "Epoch 3/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m202s\u001b[0m 77ms/step - accuracy: 0.5581 - loss: 1.5169 - val_accuracy: 0.6107 - val_loss: 1.3323\n",
      "Epoch 4/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m218s\u001b[0m 83ms/step - accuracy: 0.6257 - loss: 1.2812 - val_accuracy: 0.6371 - val_loss: 1.2361\n",
      "Epoch 5/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m218s\u001b[0m 83ms/step - accuracy: 0.6589 - loss: 1.1633 - val_accuracy: 0.6485 - val_loss: 1.2041\n",
      "Epoch 6/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m215s\u001b[0m 82ms/step - accuracy: 0.6828 - loss: 1.0728 - val_accuracy: 0.6595 - val_loss: 1.1718\n",
      "Epoch 7/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m216s\u001b[0m 83ms/step - accuracy: 0.7051 - loss: 0.9992 - val_accuracy: 0.6663 - val_loss: 1.1621\n",
      "Epoch 8/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m217s\u001b[0m 83ms/step - accuracy: 0.7217 - loss: 0.9395 - val_accuracy: 0.6689 - val_loss: 1.1660\n",
      "Epoch 9/20\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m230s\u001b[0m 88ms/step - accuracy: 0.7401 - loss: 0.8833 - val_accuracy: 0.6684 - val_loss: 1.1858\n",
      "Epoch 9: early stopping\n",
      "Restoring model weights from the end of the best epoch: 7.\n"
     ]
    }
   ],
   "source": [
    "#Forming a dictionary for class weights: each class index will be associated with a number inversely proportional with its frequency.\n",
    "#(All weights would be 1 if classes are completely balanced.)\n",
    "#This causes the model to pay more attention to underrepresented labels.  \n",
    "class_weight={}\n",
    "for i in range(31):\n",
    "    class_weight[i]=y_train.shape[0]/(y_train[y_train==i].shape[0])\n",
    "\n",
    "\n",
    "#Early stopping: the training stops if validation loss does not decrease for two epochs. Furthermore, we keep the best model.\n",
    "#https://machinelearningmastery.com/how-to-stop-training-deep-neural-networks-at-the-right-time-using-early-stopping/\n",
    "es=EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=2, restore_best_weights=True)\n",
    "\n",
    "\n",
    "opt=Adam(learning_rate=0.00005)\n",
    "model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=[\"accuracy\"])\n",
    "history = model.fit(X_train, (y_train), \n",
    "                    batch_size=32, \n",
    "                    epochs=20, validation_split=0.3,class_weight=class_weight,\n",
    "                    callbacks=[es])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "8a5b026e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "def plot_result(item):\n",
    "    plt.plot(history.history[item], label=item)\n",
    "    plt.plot(history.history[\"val_\" + item], label=\"val_\" + item)\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.locator_params(axis='x', integer=True)\n",
    "    plt.ylabel(item)\n",
    "    plt.title(\"Train and Validation {} Over Epochs\".format(item), fontsize=14)\n",
    "    plt.legend()\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "plot_result(\"loss\")\n",
    "plot_result(\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7503fbf0",
   "metadata": {},
   "source": [
    "We next evaluate the model. (Different [training APIs](https://keras.io/api/models/model_training_apis/): `compile` , `fit` , `evaluate` , `predict` etc.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "bdb59a17",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m1244/1244\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m26s\u001b[0m 21ms/step - accuracy: 0.6674 - loss: 1.1572\n",
      "Test loss: 1.158672571182251\n",
      "Test accuracy: 66.68%\n"
     ]
    }
   ],
   "source": [
    "[test_loss,test_accuracy]=model.evaluate(X_test,y_test)\n",
    "print(f'Test loss: {test_loss}')\n",
    "print(f'Test accuracy: {round(100*test_accuracy,2)}%')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "584ddd52",
   "metadata": {},
   "source": [
    "Obtaining all relevant metrics on training and test sets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ea474ea7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m3731/3731\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 20ms/step\n",
      "\u001b[1m1244/1244\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m25s\u001b[0m 20ms/step\n",
      "\n",
      "Metrics for predicting the primary arXiv category based on the cleaned text:\n",
      "\n",
      " Train Accuracy: 72.77%\n",
      " Train F1-score micro: 72.77%\n",
      " Train F1-score macro: 68.08%\n",
      " Train F1-score weighted: 72.24%\n",
      " \n",
      " Test Accuracy:  66.68%\n",
      " Test F1-score micro: 66.68%\n",
      " Test F1-score macro: 61.45%\n",
      " Test F1-score weighted: 66.07%\n"
     ]
    }
   ],
   "source": [
    "#Consists of probabilities\n",
    "y_pred_train_proba=model.predict(X_train)\n",
    "y_pred_proba=model.predict(X_test) \n",
    "\n",
    "\n",
    "#Predicted labels based on predicted probabilities.\n",
    "#https://stackoverflow.com/questions/20295046/numpy-change-max-in-each-row-to-1-all-other-numbers-to-0?rq=3\n",
    "y_pred_train=(y_pred_train_proba==y_pred_train_proba.max(axis=1)[:,None]).astype(int)\n",
    "y_pred=(y_pred_proba==y_pred_proba.max(axis=1)[:,None]).astype(int)\n",
    "\n",
    "\n",
    "print('\\nMetrics for predicting the primary arXiv category based on the cleaned text:\\n')\n",
    "print(f' Train Accuracy: {round(100*accuracy_score(y_train, y_pred_train, normalize=True),2)}%')\n",
    "print(f\" Train F1-score micro: {round(100*f1_score(y_train, y_pred_train, average='micro'),2)}%\")\n",
    "print(f\" Train F1-score macro: {round(100*f1_score(y_train, y_pred_train, average='macro'),2)}%\")\n",
    "print(f\" Train F1-score weighted: {round(100*f1_score(y_train, y_pred_train, average='weighted'),2)}%\")\n",
    "\n",
    "\n",
    "\n",
    "print(f' \\n Test Accuracy:  {round(100*accuracy_score(y_test, y_pred, normalize=True),2)}%')\n",
    "print(f\" Test F1-score micro: {round(100*f1_score(y_test, y_pred, average='micro'),2)}%\")\n",
    "print(f\" Test F1-score macro: {round(100*f1_score(y_test, y_pred, average='macro'),2)}%\")\n",
    "print(f\" Test F1-score weighted: {round(100*f1_score(y_test, y_pred, average='weighted'),2)}%\")  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "42d079a1",
   "metadata": {},
   "source": [
    "**Conclusion)** The performance of this transformer model on the test set is slightly better than what we obtained from training a CNN for the same problem of predicting the primary arXiv category. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dee62359",
   "metadata": {},
   "source": [
    "[Saving a model with custom layers](https://keras.io/guides/serialization_and_saving/#custom-objects) can be tricky. Instead, we [save the weights](https://keras.io/guides/serialization_and_saving/#model-weights-saving). Since we know the architecture, the model can be reconstructed for inference with its weights. This is the purpose of `load_transformer.py`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "8baa78f1",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Caution: A .h5 file may be generated too. \n",
    "model.save_weights('./models/clf_transformer.weights.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6e6d1a4",
   "metadata": {},
   "source": [
    "Next, we train a transformer for the multi-label problem of predicting 3-character MSC classes for which, previously, in `MSC Prediction.ipynb` we trained (and calibrated) a support vector machine:\n",
    "\n",
    "$\\hspace{3cm}$  `cleaned_text` $\\mapsto$ 3-character MSC classes (e.g. 46L, 57M, 57R) $\\hspace{1cm}$  (~500 labels).\n",
    "\n",
    "First, we delete the current model and clearing the Keras session to avoid clutter from old models/layers.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "b583844d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://stackoverflow.com/questions/50895110/what-do-i-need-k-clear-session-and-del-model-for-keras-with-tensorflow-gpu\n",
    "# https://stackoverflow.com/questions/58137677/keras-model-training-memory-leak\n",
    "\n",
    "del model            # Deleting the model. \n",
    "gc.collect()         # Clearing the remenant traces of the model with Python's garbage collection module. \n",
    "K.clear_session()    # Clearing the TensorFlow graph."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5722cb55",
   "metadata": {},
   "source": [
    "Next, we load the data and the train/validation/test split saved from before. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "ab584b68",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training instances: 104675\n",
      "Number of validation instances: 18540\n",
      "Number of test instances: 41015\n",
      "Number of labels: 531\n"
     ]
    }
   ],
   "source": [
    "data=pd.read_csv('data/data_encoded.zip',compression='zip')\n",
    "\n",
    "with open('data/train_index.json','r') as file:\n",
    "    train_index=json.load(file)\n",
    "X_train=data.iloc[train_index]['cleaned_text']\n",
    "y_train=data.iloc[train_index,1:]\n",
    "    \n",
    "with open('data/val_index.json','r') as file:\n",
    "    val_index=json.load(file)\n",
    "X_val=data.iloc[val_index]['cleaned_text']\n",
    "y_val=data.iloc[val_index,1:]\n",
    "    \n",
    "with open('data/test_index.json','r') as file:\n",
    "    test_index=json.load(file)\n",
    "X_test=data.iloc[test_index]['cleaned_text']\n",
    "y_test=data.iloc[test_index,1:]\n",
    "\n",
    "print(f'Number of training instances: {y_train.shape[0]}')\n",
    "print(f'Number of validation instances: {y_val.shape[0]}')\n",
    "print(f'Number of test instances: {y_test.shape[0]}')\n",
    "print(f'Number of labels: {y_test.shape[1]}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "757f3151",
   "metadata": {},
   "source": [
    "The training, validation and test sets were obtained from stratified sampling for multi-label data. We double check that the most common labels are distributed evenly between them. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "60f533e1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0MAAAF0CAYAAADhIOLTAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8pXeV/AAAACXBIWXMAAA9hAAAPYQGoP6dpAABhrklEQVR4nO3deVxU1f8/8NfIMsMmsiOKgDvuCmVS7gpq7plWrimmUppb5ZKJ5r6FLUqWu6Zkmrll4gKaYOFuSqZ9UEwh01JcUeH9+8Pf3C/DDMMMDqLO6/l43McDzj333vc9c+6dec+994xKRARERERERERWplRJB0BERERERFQSmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDNETadmyZVCpVAan0aNHl3R4T5zo6GioVKqSDqNAcXFxqFmzJhwcHKBSqXD06FGD9U6dOoXo6GicO3fuscaX1zfffIOYmJgS274piivGwMBA9OvXr0jLqlQqvPPOOxaNR6VSITo62qLrfJYYOu6bNWuGZs2aPfZYTHmtEhISoFKp8N133xVrLElJSYiOjsa1a9f05gUGBqJ9+/bFun1jinqM3b59G9HR0UhISLB4TEVx6dIlREdHGzyXR0dHIzAw8LHEce7cOahUKixbtuyxbI+eTbYlHQCRMUuXLkX16tV1yvz8/EoomidXZGQk2rRpU9JhGPTPP/+gd+/eaNOmDRYsWAC1Wo2qVasarHvq1ClMmjQJzZo1e2xvpvl98803+O233zB8+PAS2b4pnoYYqWQsWLCgpEMocUlJSZg0aRL69euHMmXKlHQ4FnH79m1MmjQJAEok2c3v0qVLmDRpEgIDA1GvXr0Si6Ns2bJITk5GpUqVSiwGevoxGaInWq1atRAaGmpS3fv370OlUsHW1vq6dfny5VG+fPmSDsOgP/74A/fv30evXr3QtGnTkg6HqNjcvn0bjo6OJRpDjRo1SnT7RI+TWq3GCy+8UNJh0FOOt8nRU0l7u8XKlSsxatQolCtXDmq1GmfPngUA7Ny5Ey1btkTp0qXh6OiIF198Ebt27dJbz9atW1GvXj2o1WoEBQVhzpw5ereeGLsMb+jWkDNnzuCNN96At7c31Go1goOD8cUXXxiMf82aNRg/fjz8/PxQunRptGrVCqdPn9bbzvbt29GyZUu4urrC0dERwcHBmD59ujK/oNvk4uLi0KhRIzg5OcHZ2RkRERE4cuSITp3//e9/eO211+Dn5we1Wg0fHx+0bNmywFvZ8tq0aRMaNWoER0dHuLi4oHXr1khOTlbm9+vXDy+99BIAoEePHlCpVAV+q7ls2TK8+uqrAIDmzZsrt0XmbfclS5agbt260Gg0cHd3R5cuXZCamqqznn79+sHZ2RknT55Ey5Yt4eTkBC8vL7zzzju4ffu20f1p1qwZtm7divPnz+vcmgkAzz33HF5++WWd+rVr14ZKpUJKSopStmHDBqhUKpw4cUIps2SfKChGEUGVKlUQERGht183b96Eq6sr3n77baP7n9/du3cxatQo1KtXD66urnB3d0ejRo3www8/FLjMl19+iapVq0KtVqNGjRpYu3atXp3MzEwMGjQI5cuXh729PYKCgjBp0iQ8ePDAaDy3b9/G6NGjERQUpPSB0NBQrFmzxuhy2ttu9+zZgyFDhsDT0xMeHh7o2rUrLl26pFM3NzcXs2bNQvXq1aFWq+Ht7Y0+ffrgr7/+0qnXrFkz1KpVC3v37kVYWBgcHR3Rv39/5Xwxe/ZszJw5E4GBgXBwcECzZs2ULwbGjBkDPz8/uLq6okuXLrh8+bLOuuPi4hAeHo6yZcvCwcEBwcHBGDNmDG7dumV0P7Vx5T3G+vXrV+Atx3nPXVlZWUrb2tvbo1y5chg+fLjeNrOysjBw4EB4eHjA2dkZbdq0wR9//FFoXHndvXsXI0eOhK+vLxwcHNC0aVOd89LKlSuhUql0ziVakydPhp2dnd7rphUdHY333nsPABAUFKTsa/7by7Zv344GDRrAwcEB1atXx5IlS/TWVdR+Cjz8cu7999+Hr68vHB0d8dJLL+HXX3/Vq/fPP/8gKioKNWrUgLOzM7y9vdGiRQvs27dPqXPu3Dl4eXkBACZNmqTsk/Z2u7Nnz+LNN99ElSpV4OjoiHLlyqFDhw465yDgYd+eMmUKqlWrBgcHB5QpUwZ16tTB/PnzdeoVdr5KSEjAc889BwB48803Dfan/NatW4eGDRsq72EVK1ZE//79C23HwpYz9P5cUH9XqVQ6t2AfPHgQHTt2hLu7OzQaDerXr49vv/220JjoGSRET6ClS5cKADlw4IDcv39fZxIR2bNnjwCQcuXKSbdu3WTTpk2yZcsWuXr1qqxcuVJUKpV07txZNmzYIJs3b5b27duLjY2N7Ny5U9nGzp07xcbGRl566SXZsGGDrFu3Tp577jmpUKGC5D000tLSBIAsXbpUL04AMnHiROX/kydPiqurq9SuXVtWrFghO3bskFGjRkmpUqUkOjpaqaeNPzAwUHr27Clbt26VNWvWSIUKFaRKlSry4MEDpe7XX38tKpVKmjVrJt98843s3LlTFixYIFFRUUqdiRMnSv7DeerUqaJSqaR///6yZcsW2bBhgzRq1EicnJzk5MmTSr1q1apJ5cqVZeXKlZKYmCjr16+XUaNGyZ49e4y+RqtXrxYAEh4eLhs3bpS4uDgJCQkRe3t72bdvn4iInD17Vr744gsBINOmTZPk5GSdbed1+fJlmTZtmgCQL774QpKTkyU5OVkuX74sIqLMe/3112Xr1q2yYsUKqVixori6usoff/yhrKdv375ib28vFSpUkKlTp8qOHTskOjpabG1tpX379kb36eTJk/Liiy+Kr6+vsv3k5GQRERkzZow4OzvLvXv3REQkMzNTAIiDg4NMnTpVWceQIUPEx8dHZ52W7BPGYpw/f76oVCqd9hAR5TUoqO21AgICpG/fvsr/165dk379+snKlStl9+7dsn37dhk9erSUKlVKli9frrMsAPH395caNWrImjVrZNOmTdKmTRsBIOvWrVPqZWRkiL+/vwQEBMiXX34pO3fulI8//ljUarX069dPb515j69BgwaJo6OjzJs3T/bs2SNbtmyRGTNmyGeffWZ0v7Tnk4oVK8rQoUPlp59+kq+//lrc3NykefPmOnXfeustASDvvPOObN++XWJjY8XLy0v8/f3ln3/+Ueo1bdpU3N3dxd/fXz777DPZs2ePJCYmKueLgIAA6dChg2zZskVWrVolPj4+UrVqVendu7f0799ffvzxR4mNjRVnZ2fp0KGDTgwff/yxfPLJJ7J161ZJSEiQ2NhYCQoK0ovV0HHftGlTadq0qfL/2bNndfpJcnKy9OrVSwBIXFyciIjcunVL6tWrJ56enjJv3jzZuXOnzJ8/X1xdXaVFixaSm5srIiK5ubnSvHlzUavVyrE1ceJEqVixot5rZYi2j/v7+0unTp1k8+bNsmrVKqlcubKULl1a/vzzTxERyc7OFl9fX+nZs6fO8vfv3xc/Pz959dVXC9zGhQsXZOjQoQJANmzYoOzz9evXReRhHy9fvrzUqFFDVqxYIT/99JO8+uqrAkASExOV9ZjTTw3p27evqFQqee+992THjh0yb948KVeunJQuXVrnGPv9999lyJAhsnbtWklISJAtW7bIgAEDpFSpUso5+O7du7J9+3YBIAMGDFD26ezZsyIikpiYKKNGjZLvvvtOEhMT5fvvv5fOnTuLg4OD/P7778q2pk+fLjY2NjJx4kTZtWuXbN++XWJiYnTOQ6acr65fv64cUx9++KESz4ULFwy2RVJSkqhUKnnttddk27Ztsnv3blm6dKn07t3baBuaspyh9+f8/X337t1Srlw58fX1VfrB7t27xd7eXho3bixxcXGyfft26devX4Hv9fRsYzJETyTtidbQdP/+feVNtUmTJjrL3bp1S9zd3fU+XOTk5EjdunXl+eefV8oaNmwofn5+cufOHaUsKytL3N3di5wMRURESPny5ZUTrtY777wjGo1G/v33XxH5vw8F7dq106n37bffCgDlw+2NGzekdOnS8tJLLykfSAzJ/6EoPT1dbG1tZejQoTr1bty4Ib6+vtK9e3cREbly5YoAkJiYmALXbUhOTo74+flJ7dq1JScnR2f93t7eEhYWppRp9zXvB+KCrFu3TgDoJWL//fefODg46LVXenq6qNVqeeONN5Syvn37CgCZP3++Tt2pU6cKAPn555+NxvDyyy9LQECAXvnOnTsFgOzdu1dERFatWiUuLi4SFRWl8yG1SpUqOvFYuk8YizErK0tcXFzk3Xff1SmvUaOG3gdpQ/InQ/k9ePBA7t+/LwMGDJD69evrzNMmhpmZmTr1q1evLpUrV1bKBg0aJM7OznL+/Hmd5efMmaOXsOU/vmrVqiWdO3cudD/y055P8n6BICIya9YsASAZGRkiIpKammqw3i+//CIAZNy4cUpZ06ZNBYDs2rVLp672fFG3bl2dYyMmJkYASMeOHXXqDx8+XADo9Q+t3NxcuX//viQmJgoAOXbsmDLPlGQov2+//VZUKpXOvkyfPl1KlSolKSkpOnW/++47ASDbtm0TEZEff/zR6LFlajLUoEEDnfPZuXPnxM7OTiIjI3X2zd7eXv7++2+lLC4uTi9pMWT27NkCQNLS0vTmBQQEiEaj0el/d+7cEXd3dxk0aJBSZk4/zU/bj0aMGKFTrv0CyZRjrGXLltKlSxel/J9//jGpjbXruHfvnlSpUkUnhvbt20u9evWMLmvq+SolJcXkxEHbZteuXSu0rrnLGXt/FnnYFp06dRJnZ2c5dOiQUl69enWpX7++8gWrVvv27aVs2bI6xy49+3ibHD3RVqxYgZSUFJ0p7zNBr7zyik79pKQk/Pvvv+jbty8ePHigTLm5uWjTpg1SUlJw69Yt3Lp1CykpKejatSs0Go2yvIuLCzp06FCkWO/evYtdu3ahS5cucHR01Nl+u3btcPfuXRw4cEBnmY4dO+r8X6dOHQDA+fPnlf3JyspCVFSUWaPF/fTTT3jw4AH69OmjE4dGo0HTpk2VW0bc3d1RqVIlzJ49G/PmzcORI0eQm5tb6PpPnz6NS5cuoXfv3ihV6v9OI87OznjllVdw4MCBQm9JM0dycjLu3LmjNwqTv78/WrRoYfAWyJ49e+r8/8YbbwAA9uzZU6QYXnzxRWg0GuzcuRMAEB8fj2bNmqFNmzZISkrC7du3ceHCBZw5cwatWrUCUDx9whgXFxe8+eabWLZsmXJ70+7du3Hq1Kkij/S2bt06vPjii3B2doatrS3s7OywePFivdsTAaBly5bw8fFR/rexsUGPHj1w9uxZ5TazLVu2oHnz5vDz89Npj7Zt2wIAEhMTC4zl+eefx48//ogxY8YgISEBd+7cMWtfCmtbbd/I38+ef/55BAcH6/UzNzc3tGjRwuC22rVrp3NsBAcHA4DerZba8vT0dKXsf//7H9544w34+vrCxsYGdnZ2yvN2htrdVImJiejduzd69eqFqVOnKuVbtmxBrVq1UK9ePZ3XJCIiQucWM237FHRsmeqNN97QOZ8FBAQgLCxM59gcMmQIAOCrr75Syj7//HPUrl0bTZo0MWt7+dWrVw8VKlRQ/tdoNKhatarOMfYo/bSgdurevbvBZ1pjY2PRoEEDaDQa5RjbtWuXya/1gwcPMG3aNNSoUQP29vawtbWFvb09zpw5o7OO559/HseOHUNUVBR++uknZGVl6aynKOcrU2hvqevevTu+/fZbXLx4sViXy+udd97B1q1bsW7dOjRo0ADAw9sKf//9d+X1yb+fGRkZBm9Xp2cXkyF6ogUHByM0NFRnyqts2bI6///9998AgG7dusHOzk5nmjlzJkQE//77L/777z/k5ubC19dXb5uGykxx9epVPHjwAJ999pnettu1awcAuHLlis4yHh4eOv+r1WoAUD7k/fPPPwBg9uAI2nZ47rnn9GKJi4tT4lCpVNi1axciIiIwa9YsNGjQAF5eXhg2bBhu3LhhdF8B/fYHHo72l5ubi//++8+smI0pbHva+Vq2trZ6bat9XfPXNZVGo8GLL76oJEO7du1C69at0axZM+Tk5GDfvn2Ij48HACUZKo4+UZihQ4fixo0bWL16NYCHHyDLly+PTp06mb3PGzZsQPfu3VGuXDmsWrUKycnJSElJQf/+/XH37l29+saOJ227//3339i8ebNee9SsWROAfnvk9emnn+KDDz7Axo0b0bx5c7i7u6Nz5844c+aMSftTWNua288M1dNyd3fX+d/e3t5oubY9b968icaNG+OXX37BlClTkJCQgJSUFGzYsEEnVnOdPHkSnTt3RuPGjbF48WKdeX///TeOHz+u95q4uLhARJTX5OrVq0aPLVMV1E/ytq+Pjw969OiBL7/8Ejk5OTh+/Dj27dtnkeHb88cPPOwLedv2Ufqpdj/y76ehtps3bx6GDBmChg0bYv369Thw4ABSUlLQpk0bk1/rkSNHYsKECejcuTM2b96MX375BSkpKahbt67OOsaOHYs5c+bgwIEDaNu2LTw8PNCyZUscPHhQidvc85UpmjRpgo0bNypf0JUvXx61atUq9Fm/oi6nNWXKFMTGxuLLL7/UGW1V+/44evRovf2Miooq8n7S08v6ht2iZ0r+qyWenp4AgM8++6zAEWZ8fHyUkecyMzP15ucv0145ys7O1inP/8HIzc0NNjY26N27d4EPqgcFBRnZG33ah2bzP7xdGG07fPfddwgICDBaNyAgQPlw9Mcff+Dbb79FdHQ07t27h9jYWIPLaN/QMzIy9OZdunQJpUqVgpubm1kxG1PY9rT7q/XgwQNcvXpV54OH9nU19EHIVC1btsRHH32EX3/9FX/99Rdat24NFxcXPPfcc4iPj8elS5dQtWpV+Pv7AyiePlGYypUro23btvjiiy/Qtm1bbNq0CZMmTYKNjY3Z61q1ahWCgoIQFxenc6zlPxa0jB1P2nb39PREnTp1dK5M5GVs6HwnJydMmjQJkyZNwt9//61cJerQoQN+//13k/erIHn7Wf4vIAz1s+L4ba/du3fj0qVLSEhI0Bl90dBv5pjqr7/+Qps2bVChQgWsX78ednZ2OvM9PT3h4OBgcBAB7XzgYfsYO7ZMVVA/yX9svvvuu1i5ciV++OEHbN++HWXKlNG72lJcHqWfavcjMzMT5cqVU8q1bZfXqlWr0KxZMyxcuFCn3NiXUfmtWrUKffr0wbRp03TKr1y5ojO0uK2tLUaOHImRI0fi2rVr2LlzJ8aNG4eIiAhcuHChWM9XnTp1QqdOnZCdnY0DBw5g+vTpeOONNxAYGIhGjRpZfLlly5ZhwoQJiI6O1huoQdufx44di65duxpcvlq1akXYS3paMRmiZ8qLL76IMmXKFHpbkL29PZ5//nls2LABs2fPVhKeGzduYPPmzTp1fXx8oNFocPz4cZ3y/CNqOTo6onnz5jhy5Ajq1KmjfOP7KMLCwuDq6orY2Fi89tprJn/4ioiIgK2tLf7880+9WwmNqVq1Kj788EOsX78ehw8fLrBetWrVUK5cOXzzzTcYPXq0EtetW7ewfv16ZYQ5cxV0FaRRo0ZwcHDAqlWrlBHngIcf8nbv3o1u3brprWv16tUYNmyY8v8333wDoPDf6Mj/DXFerVq1wrhx4zBhwgSUL19e+Q2sVq1aYdOmTcjMzNRp7+LoE4XFCDz8EBkeHo6+ffvCxsYGAwcOLNJ2VCoV7O3tdfpdZmZmgaPJ7dq1C3///bdyq1xOTg7i4uJQqVIlJblo3749tm3bhkqVKj1Swuzj44N+/frh2LFjiImJsciw1tpb3latWqXcogMAKSkpSE1Nxfjx4x9p/abQtrX2WND68ssvi7S+69evo23btlCpVNi2bRtKly6tV6d9+/aYNm0aPDw8jH7Ybd68OWbNmlXgsWWqNWvWYOTIkcq+nj9/HklJSejTp49OvZCQEISFhWHmzJn47bff8NZbb8HJyanQ9Zt7NdWQR+mn2nPM6tWrERISopR/++23eiPRqVQqvdf6+PHjSE5OVr5UAYzvk6F1bN26FRcvXkTlypUNxlimTBl069YNFy9exPDhw3Hu3DnUqFHD5PNVUdtYrVajadOmKFOmDH766SccOXLEaFJTlOW2b9+OgQMHon///pg4caLe/GrVqqFKlSo4duyYXgJJ1onJED1TnJ2d8dlnn6Fv3774999/0a1bN3h7e+Off/7BsWPH8M8//yjfwH388cdo06YNWrdujVGjRiEnJwczZ86Ek5MT/v33X2WdKpUKvXr1wpIlS1CpUiXUrVsXv/76q8EPAPPnz8dLL72Exo0bY8iQIQgMDMSNGzdw9uxZbN68Gbt37zZ7f+bOnYvIyEi0atUKAwcOhI+PD86ePYtjx47h888/N7hcYGAgJk+ejPHjx+N///sf2rRpAzc3N/z999/49ddflW/Yjx8/jnfeeQevvvoqqlSpAnt7e+zevRvHjx/HmDFjCoyrVKlSmDVrFnr27In27dtj0KBByM7OxuzZs3Ht2jXMmDHDrP3UqlWrFgBg0aJFcHFxgUajQVBQEDw8PDBhwgSMGzcOffr0weuvv46rV69i0qRJ0Gg0em949vb2mDt3Lm7evInnnnsOSUlJmDJlCtq2basM9V2Q2rVrY8OGDVi4cCFCQkJQqlQp5fbMkJAQuLm5YceOHXjzzTeVZVq1aoWPP/5Y+TsvS/eJwmIEgNatW6NGjRrYs2cPevXqBW9vb7O3ATz8QLhhwwZERUWhW7duuHDhAj7++GOULVvW4K1pnp6eaNGiBSZMmAAnJycsWLAAv//+u87w2pMnT0Z8fDzCwsIwbNgwVKtWDXfv3sW5c+ewbds2xMbGFnhbaMOGDdG+fXvUqVMHbm5uSE1NxcqVK4ucfOdXrVo1vPXWW/jss89QqlQptG3bFufOncOECRPg7++PESNGPPI2ChMWFgY3NzcMHjwYEydOhJ2dHVavXo1jx44VaX1vvPEGTp06hUWLFuHChQu4cOGCMk/7+2TDhw/H+vXr0aRJE4wYMQJ16tRBbm4u0tPTsWPHDowaNQoNGzZEeHg4mjRpgvfffx+3bt1CaGgo9u/fj5UrV5oV0+XLl9GlSxcMHDgQ169fx8SJE6HRaDB27Fi9uu+++64yLL/2FqbC1K5dG8DDY69v376ws7NDtWrV4OLiYnKMj9JPg4OD0atXL8TExMDOzg6tWrXCb7/9hjlz5uglo+3bt8fHH3+MiRMnomnTpjh9+jQmT56MoKAgncTJxcUFAQEB+OGHH9CyZUu4u7vD09MTgYGBaN++PZYtW4bq1aujTp06OHToEGbPnq0XX4cOHZTf7vPy8sL58+cRExODgIAAVKlSRWkzU85XlSpVgoODA1avXo3g4GA4OzvDz8/P4BWzjz76CH/99RdatmyJ8uXL49q1a5g/f77Os3CGFGW5tLQ0vPrqq6hYsSLefPNNvWec6tevD7VajS+//BJt27ZFREQE+vXrh3LlyuHff/9FamoqDh8+jHXr1hUYFz2DSnoEByJDtKM/5R/dSKuwEcoSExPl5ZdfFnd3d7Gzs5Ny5crJyy+/rFd/06ZNUqdOHWUo5hkzZhgcoen69esSGRkpPj4+4uTkJB06dJBz584ZHN0nLS1N+vfvL+XKlRM7Ozvx8vKSsLAwmTJlSqHxFzQyzrZt26Rp06bi5OQkjo6OUqNGDZk5c6Yy31DMIiIbN26U5s2bS+nSpUWtVktAQIB069ZNGWL877//ln79+kn16tXFyclJnJ2dpU6dOvLJJ5/oDO9dkI0bN0rDhg1Fo9GIk5OTtGzZUvbv369Tx5zR5EQejroVFBQkNjY2em3x9ddfK6+Xq6urdOrUSW9Up759+4qTk5McP35cmjVrJg4ODuLu7i5DhgyRmzdvFrr9f//9V7p16yZlypQRlUql165dunQRALJ69Wql7N69e+Lk5CSlSpWS//77T2+dlu4ThcUoIhIdHa0MT28qQ6PJzZgxQwIDA0WtVktwcLB89dVXBvsbAHn77bdlwYIFUqlSJbGzs5Pq1avrtJPWP//8I8OGDZOgoCCxs7MTd3d3CQkJkfHjx+u8RvmPrzFjxkhoaKi4ubmJWq2WihUryogRI+TKlStG96ug84m2zfOOXpiTkyMzZ86UqlWrip2dnXh6ekqvXr30hg1u2rSp1KxZU29b2tdr9uzZBreV//U1FFtSUpI0atRIHB0dxcvLSyIjI+Xw4cN6/cCU0eQCAgIKHJkzb9vevHlTPvzwQ6lWrZpyfNWuXVtGjBihM0LgtWvXpH///lKmTBlxdHSU1q1by++//27WaHIrV66UYcOGiZeXl6jVamncuLEcPHjQ4DLZ2dmiVqulTZs2Rted39ixY8XPz09KlSql8xoHBATIyy+/rFff0Ch8pvbTguIeNWqUeHt7i0ajkRdeeEGSk5P1jrHs7GwZPXq0lCtXTjQajTRo0EA2btwoffv21RsxcufOnVK/fn1Rq9U6o9L9999/MmDAAPH29hZHR0d56aWXZN++fXr7NHfuXAkLCxNPT0/lPW/AgAFy7tw5ne2Ycr4SEVmzZo1Ur15d7OzsjL7+W7ZskbZt20q5cuXE3t5evL29pV27dspPMBTElOXynx+1faygKe8Ig8eOHZPu3buLt7e32NnZia+vr7Ro0UJiY2ONxkXPHpWISPGkWURPp+joaEyaNAk8NJ5O/fr1w3fffYebN2+WdCglKjQ0VO8HYYmeNps3b0bHjh2xdetW5SF+IiJL4m1yRETPiKysLPz222/YsmULDh06hO+//76kQyIqklOnTuH8+fMYNWoU6tWrpwxpTURkaUyGiIieEYcPH0bz5s3h4eGBiRMnonPnziUdElGRREVFYf/+/WjQoAGWL19eLCP3EREBAG+TIyIiIiIiq8QfXSUiIiIiIqvEZIiIiIiIiKwSkyEiIiIiIrJKz8wACrm5ubh06RJcXFz4oCURERERkRUTEdy4cQN+fn4oVarg6z/PTDJ06dIl+Pv7l3QYRERERET0hLhw4QLKly9f4PxnJhlycXEB8HCHS5cuXcLREBERERFRScnKyoK/v7+SIxTkmUmGtLfGlS5dmskQEREREREV+vgMB1AgIiIiIiKrxGSIiIiIiIisEpMhIiIiIiKySs/MM0NERERE9HTKycnB/fv3SzoMeorY2dnBxsbmkdfDZIiIiIiISoSIIDMzE9euXSvpUOgpVKZMGfj6+j7Sb4wyGSIiIiKiEqFNhLy9veHo6PhIH2rJeogIbt++jcuXLwMAypYtW+R1MRkiIiIioscuJydHSYQ8PDxKOhx6yjg4OAAALl++DG9v7yLfMscBFIiIiIjosdM+I+To6FjCkdDTStt3HuV5MyZDRERERFRieGscFZUl+g6TISIiIiIiskpMhoiIiIiISlizZs0wfPhwk+ufO3cOKpUKR48eLbaYrAEHUCAiIiKiJ8vjvnVOxOSqhd2a1bdvXyxbtszsEDZs2AA7OzuT6/v7+yMjIwOenp5mb+tx6tevH65du4aNGzeWdCgGMRkqLsYOFCMHnGqS8QNMJpp+sBIRERGRZWVkZCh/x8XF4aOPPsLp06eVMu0oZ1r37983Kclxd3c3Kw4bGxv4+vqatQzp421yJUClKngiIiIioieXr6+vMrm6ukKlUin/3717F2XKlMG3336LZs2aQaPRYNWqVbh69Spef/11lC9fHo6OjqhduzbWrFmjs978t8kFBgZi2rRp6N+/P1xcXFChQgUsWrRImZ//NrmEhASoVCrs2rULoaGhcHR0RFhYmE6iBgBTpkyBt7c3XFxcEBkZiTFjxqBevXoF7u9///2Hnj17wsvLCw4ODqhSpQqWLl2qzL948SJ69OgBNzc3eHh4oFOnTjh37hwAIDo6GsuXL8cPP/wAlUoFlUqFhISEIrV7cWEyRERERERkQR988AGGDRuG1NRURERE4O7duwgJCcGWLVvw22+/4a233kLv3r3xyy+/GF3P3LlzERoaiiNHjiAqKgpDhgzB77//bnSZ8ePHY+7cuTh48CBsbW3Rv39/Zd7q1asxdepUzJw5E4cOHUKFChWwcOFCo+ubMGECTp06hR9//BGpqalYuHChcmve7du30bx5czg7O2Pv3r34+eef4ezsjDZt2uDevXsYPXo0unfvjjZt2iAjIwMZGRkICwszsRUfD94mR0RERERkQcOHD0fXrl11ykaPHq38PXToUGzfvh3r1q1Dw4YNC1xPu3btEBUVBeBhgvXJJ58gISEB1atXL3CZqVOnomnTpgCAMWPG4OWXX8bdu3eh0Wjw2WefYcCAAXjzzTcBAB999BF27NiBmzdvFri+9PR01K9fH6GhoQAeXrHSWrt2LUqVKoWvv/5aeZZq6dKlKFOmDBISEhAeHg4HBwdkZ2c/sbf08coQEREREZEFaRMHrZycHEydOhV16tSBh4cHnJ2dsWPHDqSnpxtdT506dZS/tbfjXb582eRlypYtCwDKMqdPn8bzzz+vUz////kNGTIEa9euRb169fD+++8jKSlJmXfo0CGcPXsWLi4ucHZ2hrOzM9zd3XH37l38+eefRtf7pOCVISIiIiIiC3JyctL5f+7cufjkk08QExOD2rVrw8nJCcOHD8e9e/eMrif/wAsqlQq5ubkmL6O9WpN3mfyj4UkhI+m1bdsW58+fx9atW7Fz5060bNkSb7/9NubMmYPc3FyEhIRg9erVest5eXkZXe+TgleGiIiIiIiK0b59+9CpUyf06tULdevWRcWKFXHmzJnHHke1atXw66+/6pQdPHiw0OW8vLzQr18/rFq1CjExMcpADg0aNMCZM2fg7e2NypUr60yurq4AAHt7e+Tk5Fh+ZyyEyRARERERUTGqXLky4uPjkZSUhNTUVAwaNAiZmZmPPY6hQ4di8eLFWL58Oc6cOYMpU6bg+PHjRn876aOPPsIPP/yAs2fP4uTJk9iyZQuCg4MBAD179oSnpyc6deqEffv2IS0tDYmJiXj33Xfx119/AXj4jNHx48dx+vRpXLlyBffv338s+2oqJkNERERERMVowoQJaNCgASIiItCsWTP4+vqic+fOjz2Onj17YuzYsRg9ejQaNGiAtLQ09OvXDxqNpsBl7O3tMXbsWNSpUwdNmjSBjY0N1q5dCwBwdHTE3r17UaFCBXTt2hXBwcHo378/7ty5g9KlSwMABg4ciGrVqiE0NBReXl7Yv3//Y9lXU6mksBsFnxJZWVlwdXXF9evXlcYvUUYybBWMNHk0f3SViIiInn13795FWloagoKCjH4Yp+LVunVr+Pr6YuXKlSUditmM9SFTcwMOoEBEREREZAVu376N2NhYREREwMbGBmvWrMHOnTsRHx9f0qGVGCZDRERERERWQKVSYdu2bZgyZQqys7NRrVo1rF+/Hq1atSrp0EoMkyEiIiIiIivg4OCAnTt3lnQYTxQOoEBERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERESPWbNmzTB8+HDl/8DAQMTExBhdRqVSYePGjY+8bUut51nAZIiIiIiInigq1eOdzNGhQ4cCf6Q0OTkZKpUKhw8fNnufU1JS8NZbb5m9nDHR0dGoV6+eXnlGRgbatm1r0W1ZminJoSUwGSIiIiIiMtGAAQOwe/dunD9/Xm/ekiVLUK9ePTRo0MDs9Xp5ecHR0dESIRbK19cXarX6sWzrScdkyFoU8esQ1SRVgRMRERGRtWnfvj28vb2xbNkynfLbt28jLi4OAwYMwNWrV/H666+jfPnycHR0RO3atbFmzRqj681/JeTMmTNo0qQJNBoNatSogfj4eL1lPvjgA1StWhWOjo6oWLEiJkyYgPv37wMAli1bhkmTJuHYsWNQqVRQqVRKzPlvkztx4gRatGgBBwcHeHh44K233sLNmzeV+f369UPnzp0xZ84clC1bFh4eHnj77beVbRly7NgxNG/eHC4uLihdujRCQkJw8OBBZX5SUhKaNGkCBwcH+Pv7Y9iwYbh16xaAh7cQnj9/HiNGjFBiLy5Mhshil42JiIiInnW2trbo06cPli1bBhFRytetW4d79+6hZ8+euHv3LkJCQrBlyxb89ttveOutt9C7d2/88ssvJm0jNzcXXbt2hY2NDQ4cOIDY2Fh88MEHevVcXFywbNkynDp1CvPnz8dXX32FTz75BADQo0cPjBo1CjVr1kRGRgYyMjLQo0cPvXXcvn0bbdq0gZubG1JSUrBu3Trs3LkT77zzjk69PXv24M8//8SePXuwfPlyLFu2TC8hzKtnz54oX748UlJScOjQIYwZMwZ2dnYAHiZfERER6Nq1K44fP464uDj8/PPPyjY3bNiA8uXLY/LkyUrsxcW22NZMRERERPQM6t+/P2bPno2EhAQ0b94cwMNb5Lp27Qo3Nze4ublh9OjRSv2hQ4di+/btWLduHRo2bFjo+nfu3InU1FScO3cO5cuXBwBMmzZN7zmfDz/8UPk7MDAQo0aNQlxcHN5//304ODjA2dkZtra28PX1LXBbq1evxp07d7BixQo4OTkBAD7//HN06NABM2fOhI+PDwDAzc0Nn3/+OWxsbFC9enW8/PLL2LVrFwYOHGhwvenp6XjvvfdQvXp1AECVKlWUebNnz8Ybb7yhDCBRpUoVfPrpp2jatCkWLlwId3d32NjYwMXFxWjslsBkiIiIiIjIDNWrV0dYWBiWLFmC5s2b488//8S+ffuwY8cOAEBOTg5mzJiBuLg4XLx4EdnZ2cjOzlaSjcKkpqaiQoUKSiIEAI0aNdKr99133yEmJgZnz57FzZs38eDBA5QuXdqsfUlNTUXdunV1YnvxxReRm5uL06dPK8lQzZo1YWNjo9QpW7YsTpw4UeB6R44cicjISKxcuRKtWrXCq6++ikqVKgEADh06hLNnz2L16tVKfRFBbm4u0tLSEBwcbNY+PAreJkdEREREZKYBAwZg/fr1yMrKwtKlSxEQEICWLVsCAObOnYtPPvkE77//Pnbv3o2jR48iIiIC9+7dM2ndeW+/08r/3MyBAwfw2muvoW3bttiyZQuOHDmC8ePHm7yNvNsq6JmcvOXaW9zyzsvNzS1wvdHR0Th58iRefvll7N69GzVq1MD3338P4OFtgIMGDcLRo0eV6dixYzhz5oySMD0uvDJERERERGSm7t27491338U333yD5cuXY+DAgUrysG/fPnTq1Am9evUC8PDD/5kzZ0y+4lGjRg2kp6fj0qVL8PPzA/Bw2O689u/fj4CAAIwfP14pyz/Cnb29PXJycgrd1vLly3Hr1i3l6tD+/ftRqlQpVK1a1aR4C1K1alVUrVoVI0aMwOuvv46lS5eiS5cuaNCgAU6ePInKlSsXuKwpsVsCrwwREREREZnJ2dkZPXr0wLhx43Dp0iX069dPmVe5cmXEx8cjKSkJqampGDRoEDIzM01ed6tWrVCtWjX06dMHx44dw759+3SSHu020tPTsXbtWvz555/49NNPlSsvWoGBgUhLS8PRo0dx5coVZGdn622rZ8+e0Gg06Nu3L3777Tfs2bMHQ4cORe/evZVb5Mx1584dvPPOO0hISMD58+exf/9+pKSkKMngBx98gOTkZLz99ts4evQozpw5g02bNmHo0KE6se/duxcXL17ElStXihSHKZgMEREREREVwYABA/Dff/+hVatWqFChglI+YcIENGjQABEREWjWrBl8fX3RuXNnk9dbqlQpfP/998jOzsbzzz+PyMhITJ06VadOp06dMGLECLzzzjuoV68ekpKSMGHCBJ06r7zyCtq0aYPmzZvDy8vL4PDejo6O+Omnn/Dvv//iueeeQ7du3dCyZUt8/vnn5jVGHjY2Nrh69Sr69OmDqlWronv37mjbti0mTZoEAKhTpw4SExNx5swZNG7cGPXr18eECRNQtmxZZR2TJ0/GuXPnUKlSJXh5eRU5lsKoxNBNiU+hrKwsuLq64vr162Y/OFYsjIxLrYKRJo82Pp61TCziy1UM8RQ5FiIiIrJ6d+/eRVpaGoKCgqDRaEo6HHoKGetDpuYGvDJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFapSMnQggULlF96DQkJwb59+4zWT0xMREhICDQaDSpWrIjY2NgC665duxYqlQqdO3cuSmhEREREREQmsTV3gbi4OAwfPhwLFizAiy++iC+//BJt27bFqVOnUKFCBb36aWlpaNeuHQYOHIhVq1Zh//79iIqKgpeXF1555RWduufPn8fo0aPRuHHjou8RERERET3VVJNUj3V7MlFMrqtSGY+tb9++WLZsWZHiCAwMxPDhwzF8+PAiLW+KZcuWYfjw4bh27VqxbeNpYvaVoXnz5mHAgAGIjIxEcHAwYmJi4O/vj4ULFxqsHxsbiwoVKiAmJgbBwcGIjIxE//79MWfOHJ16OTk56NmzJyZNmoSKFSsWbW+IiIiIiIpRRkaGMsXExKB06dI6ZfPnzy/pEMkMZiVD9+7dw6FDhxAeHq5THh4ejqSkJIPLJCcn69WPiIjAwYMHcf/+faVs8uTJ8PLywoABA0yKJTs7G1lZWToTEREREVFx8vX1VSZXV1eoVCqdsr179+o8HjJp0iQ8ePBAWT46OhoVKlSAWq2Gn58fhg0bBgBo1qwZzp8/jxEjRkClUhm9AlXQOoCHn9fff/99lCtXDk5OTmjYsCESEhIAAAkJCXjzzTdx/fp1ZRvR0dHF0k5PC7Nuk7ty5QpycnLg4+OjU+7j44PMzEyDy2RmZhqs/+DBA1y5cgVly5bF/v37sXjxYhw9etTkWKZPn45JkyaZEz4RERERUbH56aef0KtXL3z66ado3Lgx/vzzT7z11lsAgIkTJ+K7777DJ598grVr16JmzZrIzMzEsWPHAAAbNmxA3bp18dZbb2HgwIEFbsPYOgDgzTffxLlz57B27Vr4+fnh+++/R5s2bXDixAmEhYUhJiYGH330EU6fPg0AcHZ2LsYWefKZ/cwQoH+vpIgYzV4N1deW37hxA7169cJXX30FT09Pk2MYO3YsRo4cqfyflZUFf39/k5cnIiIiIrKkqVOnYsyYMejbty8AoGLFivj444/x/vvvY+LEiUhPT4evry9atWoFOzs7VKhQAc8//zwAwN3dHTY2NnBxcYGvr2+B2zC2jj///BNr1qzBX3/9BT8/PwDA6NGjsX37dixduhTTpk3TuZpFZiZDnp6esLGx0bsKdPnyZb2rP1q+vr4G69va2sLDwwMnT57EuXPn0KFDB2V+bm7uw+BsbXH69GlUqlRJb71qtRpqtdqc8ImIiIiIis2hQ4eQkpKCqVOnKmU5OTm4e/cubt++jVdffRUxMTGoWLEi2rRpg3bt2qFDhw6wtTX9I7mxdRw+fBgigqpVq+osk52dDQ8PD4vt57PErGTI3t4eISEhiI+PR5cuXZTy+Ph4dOrUyeAyjRo1wubNm3XKduzYgdDQUNjZ2aF69eo4ceKEzvwPP/wQN27cwPz583m1h4iIiIieCrm5uZg0aRK6du2qN0+j0cDf3x+nT59GfHw8du7ciaioKMyePRuJiYmws7MzaRvG1pGbmwsbGxscOnQINjY2OstZ++1wBTH7NrmRI0eid+/eCA0NRaNGjbBo0SKkp6dj8ODBAB7evnbx4kWsWLECADB48GB8/vnnGDlyJAYOHIjk5GQsXrwYa9asAfCwY9SqVUtnG2XKlAEAvXIiIiIioidVgwYNcPr0aVSuXLnAOg4ODujYsSM6duyIt99+W7kw0KBBA9jb2yMnJ6fQ7RS0jvr16yMnJweXL18u8KdqTN2GtTA7GerRoweuXr2KyZMnIyMjA7Vq1cK2bdsQEBAA4OFwg+np6Ur9oKAgbNu2DSNGjMAXX3wBPz8/fPrpp3q/MURERERE9DT76KOP0L59e/j7++PVV19FqVKlcPz4cZw4cQJTpkzBsmXLkJOTg4YNG8LR0RErV66Eg4OD8jk6MDAQe/fuxWuvvQa1Wm3weXpj6/Dw8EDPnj3Rp08fzJ07F/Xr18eVK1ewe/du1K5dG+3atUNgYCBu3ryJXbt2oW7dunB0dISjo+PjbqonRpEGUIiKikJUVJTBeYZ+ZKpp06Y4fPiwyesv6g9VERERERGVlIiICGzZsgWTJ0/GrFmzlEdCIiMjATy8+2nGjBkYOXIkcnJyULt2bWzevFl5nmfy5MkYNGgQKlWqhOzsbGXQsbwKW8fSpUsxZcoUjBo1ChcvXoSHhwcaNWqEdu3aAQDCwsIwePBg5QLHxIkTrXp4bZUYauWnUFZWFlxdXXH9+nWULl26pMMBjI2uByNNHm38V43N+YXk4o6nyLEQERGR1bt79y7S0tIQFBQEjUZT0uHQU8hYHzI1NzDrR1eJiIiIiIieFUyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBKTISIiIiIqMc/IWF5UAizRd4o0tDZRSVFN4uh2REREzwI7OzsAwO3bt+Hg4FDC0dDT6Pbt2wD+ry8VBZMhKhlFHurb8qEQERHR42djY4MyZcrg8uXLAABHR0eojHw+INISEdy+fRuXL19GmTJlYGNjU+R1MRkiIiIiohLh6+sLAEpCRGSOMmXKKH2oqJgMEREREVGJUKlUKFu2LLy9vXH//v2SDoeeInZ2do90RUiLyRARERERlSgbGxuLfLAlMhdHkyMiIiIiIqvEZIiIiIiIiKwSkyEiIiIiIrJKTIaIiIiIiMgqMRkiIiIiIiKrxGSIiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBKTISIiIiIiskpMhoiIiIiIyCoxGSIiIiIiIqvEZIiIiIiIiKwSkyEiIiIiIrJKTIaIiIiIiMgqMRkiIiIiIiKrxGSIiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKtmWdABETwSVquB5IgUvNqng5WRiwcsRERERUcnjlSGiQqhUBU9ERERE9PRiMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSXbkg6AiAxQqQqeJ1LwYpMKXk4mFrwcERERkTXilSGip4xKVfBERERERKZjMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWiUNrE1HhjAxVp4KRIbujOdQ3ERERPbmKdGVowYIFCAoKgkajQUhICPbt22e0fmJiIkJCQqDRaFCxYkXExsbqzN+wYQNCQ0NRpkwZODk5oV69eli5cmVRQiMiIiIiIjKJ2clQXFwchg8fjvHjx+PIkSNo3Lgx2rZti/T0dIP109LS0K5dOzRu3BhHjhzBuHHjMGzYMKxfv16p4+7ujvHjxyM5ORnHjx/Hm2++iTfffBM//fRT0feMiIiIiIjICLOToXnz5mHAgAGIjIxEcHAwYmJi4O/vj4ULFxqsHxsbiwoVKiAmJgbBwcGIjIxE//79MWfOHKVOs2bN0KVLFwQHB6NSpUp49913UadOHfz8889F3zMiIiIiIiIjzEqG7t27h0OHDiE8PFynPDw8HElJSQaXSU5O1qsfERGBgwcP4v79+3r1RQS7du3C6dOn0aRJkwJjyc7ORlZWls5ERERERERkKrOSoStXriAnJwc+Pj465T4+PsjMzDS4TGZmpsH6Dx48wJUrV5Sy69evw9nZGfb29nj55Zfx2WefoXXr1gXGMn36dLi6uiqTv7+/ObtCRERERERWrkgDKKjyjSwlInplhdXPX+7i4oKjR48iJSUFU6dOxciRI5GQkFDgOseOHYvr168r04ULF4qwJ0REREREZK3MGlrb09MTNjY2eleBLl++rHf1R8vX19dgfVtbW3h4eChlpUqVQuXKlQEA9erVQ2pqKqZPn45mzZoZXK9arYZarTYnfCIiIiIiIoVZV4bs7e0REhKC+Ph4nfL4+HiEhYUZXKZRo0Z69Xfs2IHQ0FDY2dkVuC0RQXZ2tjnhERERERERmczsH10dOXIkevfujdDQUDRq1AiLFi1Ceno6Bg8eDODh7WsXL17EihUrAACDBw/G559/jpEjR2LgwIFITk7G4sWLsWbNGmWd06dPR2hoKCpVqoR79+5h27ZtWLFiRYEj1BERERERET0qs5OhHj164OrVq5g8eTIyMjJQq1YtbNu2DQEBAQCAjIwMnd8cCgoKwrZt2zBixAh88cUX8PPzw6effopXXnlFqXPr1i1ERUXhr7/+goODA6pXr45Vq1ahR48eFthFIiIiIiIifSrRjmbwlMvKyoKrqyuuX7+O0qVLl3Q4gLEBJWCkyaMLXg4AZGIRX65iiKfIsTAexvOo8RAREREZYWpuUKTR5IiIiIiIiJ52TIaIiIiIiMgqMRkiIiIiIiKrxGSIiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBKTISIiIiIiskpMhoiIiIiIyCoxGSIiIiIiIqvEZIiIiIiIiKwSkyEiIiIiIrJKTIaIiIiIiMgqMRkiIiIiIiKrxGSIiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBKTISIiIiIiskpMhoiIiIiIyCrZlnQARERmU6kKngUpeLnogpeTiUaWIyIiomcSrwwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVeJockREj4qj2xERET2VeGWIiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBJHkyMietZwdDsiIiKTMBkiIqLiYyQxAwBIwUmWahKTMyIiKl68TY6IiEqMSlXwREREVNyYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWib8zRERE1oO/e0RERHkwGSIiIvr/jOZK0Y97g2ByRkRUzHibHBER0ROKP0pLRFS8mAwREREREZFV4m1yRERE9JCRS04qFHLrXTRv2yOipw+vDBERERERkVViMkRERERERFaJyRAREREREVmlIiVDCxYsQFBQEDQaDUJCQrBv3z6j9RMTExESEgKNRoOKFSsiNjZWZ/5XX32Fxo0bw83NDW5ubmjVqhV+/fXXooRGRERERERkErMHUIiLi8Pw4cOxYMECvPjii/jyyy/Rtm1bnDp1ChUqVNCrn5aWhnbt2mHgwIFYtWoV9u/fj6ioKHh5eeGVV14BACQkJOD1119HWFgYNBoNZs2ahfDwcJw8eRLlypV79L0kIiKip4+xMcSN/AYTwN9hIiLTmH1laN68eRgwYAAiIyMRHByMmJgY+Pv7Y+HChQbrx8bGokKFCoiJiUFwcDAiIyPRv39/zJkzR6mzevVqREVFoV69eqhevTq++uor5ObmYteuXUXfMyIiInpmGfsNJv4OExGZyqxk6N69ezh06BDCw8N1ysPDw5GUlGRwmeTkZL36EREROHjwIO7fv29wmdu3b+P+/ftwd3c3JzwiIiIiIiKTmXWb3JUrV5CTkwMfHx+dch8fH2RmZhpcJjMz02D9Bw8e4MqVKyhbtqzeMmPGjEG5cuXQqlWrAmPJzs5Gdna28n9WVpY5u0JERERERFauSD+6qsp3/VlE9MoKq2+oHABmzZqFNWvWICEhARqNpsB1Tp8+HZMmTTInbCIiIqKi4zNMRM8cs26T8/T0hI2Njd5VoMuXL+td/dHy9fU1WN/W1hYeHh465XPmzMG0adOwY8cO1KlTx2gsY8eOxfXr15XpwoUL5uwKERERkcXwGSaip5NZyZC9vT1CQkIQHx+vUx4fH4+wsDCDyzRq1Eiv/o4dOxAaGgo7OzulbPbs2fj444+xfft2hIaGFhqLWq1G6dKldSYiIiIiIiJTmT2a3MiRI/H1119jyZIlSE1NxYgRI5Ceno7BgwcDeHjFpk+fPkr9wYMH4/z58xg5ciRSU1OxZMkSLF68GKNHj1bqzJo1Cx9++CGWLFmCwMBAZGZmIjMzEzdv3rTALhIREREREekz+5mhHj164OrVq5g8eTIyMjJQq1YtbNu2DQEBAQCAjIwMpKenK/WDgoKwbds2jBgxAl988QX8/Pzw6aefKr8xBDz8Edd79+6hW7duOtuaOHEioqOji7hrREREREREBSvSAApRUVGIiooyOG/ZsmV6ZU2bNsXhw4cLXN+5c+eKEgYREREREVGRmX2bHBERERER0bOAyRAREREREVklJkNERERERGSVivTMEBERERGVMP4ILNEj45UhIiIiomcMfwSWyDRMhoiIiIiIyCoxGSIiIiIiIqvEZ4aIiIiI6NEV8RkmY88vAXyGiYoXkyEiIiIiKlZGn1OKfvwbVcFIghXN5MyaMBkiIiIiIipuTM6eSHxmiIiIiIiIrBKTISIiIiIiskpMhoiIiIiIyCoxGSIiIiIiIqvEZIiIiIiIiKwSR5MjIiIiIrI2xTC63dM4sh2vDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWKlAwtWLAAQUFB0Gg0CAkJwb59+4zWT0xMREhICDQaDSpWrIjY2Fid+SdPnsQrr7yCwMBAqFQqxMTEFCUsIiIiIiIik5mdDMXFxWH48OEYP348jhw5gsaNG6Nt27ZIT083WD8tLQ3t2rVD48aNceTIEYwbNw7Dhg3D+vXrlTq3b99GxYoVMWPGDPj6+hZ9b4iIiIiIiExkdjI0b948DBgwAJGRkQgODkZMTAz8/f2xcOFCg/VjY2NRoUIFxMTEIDg4GJGRkejfvz/mzJmj1Hnuuecwe/ZsvPbaa1Cr1UXfGyIiIiIiIhOZlQzdu3cPhw4dQnh4uE55eHg4kpKSDC6TnJysVz8iIgIHDx7E/fv3zQz3/2RnZyMrK0tnIiIiIiIiMpVZydCVK1eQk5MDHx8fnXIfHx9kZmYaXCYzM9Ng/QcPHuDKlStmhvt/pk+fDldXV2Xy9/cv8rqIiIiIiMj6FGkABZVKpfO/iOiVFVbfULk5xo4di+vXryvThQsXirwuIiIiIiKyPrbmVPb09ISNjY3eVaDLly/rXf3R8vX1NVjf1tYWHh4eZob7f9RqNZ8vIiIiIiKiIjPrypC9vT1CQkIQHx+vUx4fH4+wsDCDyzRq1Eiv/o4dOxAaGgo7OzszwyUiIiIiIrIMs2+TGzlyJL7++mssWbIEqampGDFiBNLT0zF48GAAD29f69Onj1J/8ODBOH/+PEaOHInU1FQsWbIEixcvxujRo5U69+7dw9GjR3H06FHcu3cPFy9exNGjR3H27FkL7CIREREREZE+s26TA4AePXrg6tWrmDx5MjIyMlCrVi1s27YNAQEBAICMjAyd3xwKCgrCtm3bMGLECHzxxRfw8/PDp59+ildeeUWpc+nSJdSvX1/5f86cOZgzZw6aNm2KhISER9g9IiIiIiIiw8xOhgAgKioKUVFRBuctW7ZMr6xp06Y4fPhwgesLDAxUBlUgIiIiIiJ6HIo0mhwREREREdHTjskQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklZgMERERERGRVWIyREREREREVonJEBERERERWSUmQ0REREREZJWYDBERERERkVViMkRERERERFaJyRAREREREVklJkNERERERGSVmAwREREREZFVYjJERERERERWickQERERERFZJSZDRERERERklYqUDC1YsABBQUHQaDQICQnBvn37jNZPTExESEgINBoNKlasiNjYWL0669evR40aNaBWq1GjRg18//33RQmNiIiIiIjIJGYnQ3FxcRg+fDjGjx+PI0eOoHHjxmjbti3S09MN1k9LS0O7du3QuHFjHDlyBOPGjcOwYcOwfv16pU5ycjJ69OiB3r1749ixY+jduze6d++OX375peh7RkREREREZITZydC8efMwYMAAREZGIjg4GDExMfD398fChQsN1o+NjUWFChUQExOD4OBgREZGon///pgzZ45SJyYmBq1bt8bYsWNRvXp1jB07Fi1btkRMTEyRd4yIiIiIiMgYW3Mq37t3D4cOHcKYMWN0ysPDw5GUlGRwmeTkZISHh+uURUREYPHixbh//z7s7OyQnJyMESNG6NUxlgxlZ2cjOztb+f/69esAgKysLHN2qYQYifFuIUsWy/4VLZ7ia2vGYxzjMY7xGMd4jGM8RtZqfDbjMT6b8RQ86yn67AMwnpI5F5pPG4uIGK8oZrh48aIAkP379+uUT506VapWrWpwmSpVqsjUqVN1yvbv3y8A5NKlSyIiYmdnJ6tXr9aps3r1arG3ty8wlokTJwoATpw4ceLEiRMnTpw4cTI4XbhwwWh+Y9aVIS2VSqXzv4jolRVWP3+5uescO3YsRo4cqfyfm5uLf//9Fx4eHkaXe9JkZWXB398fFy5cQOnSpUs6HMbDeJ6JWBgP42E8jIfxMB7G82zGYyoRwY0bN+Dn52e0nlnJkKenJ2xsbJCZmalTfvnyZfj4+BhcxtfX12B9W1tbeHh4GK1T0DoBQK1WQ61W65SVKVPG1F154pQuXfqJ6mCMxzjGU7AnKRaA8RSG8RjHeIxjPMYxHuMYj3GM59G5uroWWsesARTs7e0REhKC+Ph4nfL4+HiEhYUZXKZRo0Z69Xfs2IHQ0FDY2dkZrVPQOomIiIiIiB6V2bfJjRw5Er1790ZoaCgaNWqERYsWIT09HYMHDwbw8Pa1ixcvYsWKFQCAwYMH4/PPP8fIkSMxcOBAJCcnY/HixVizZo2yznfffRdNmjTBzJkz0alTJ/zwww/YuXMnfv75ZwvtJhERERERkS6zk6EePXrg6tWrmDx5MjIyMlCrVi1s27YNAQEBAICMjAyd3xwKCgrCtm3bMGLECHzxxRfw8/PDp59+ildeeUWpExYWhrVr1+LDDz/EhAkTUKlSJcTFxaFhw4YW2MUnm1qtxsSJE/Vu+SspjMc4xvN0xAIwnsIwHuMYj3GMxzjGYxzjMY7xPF4qkcLGmyMiIiIiInr2mP2jq0RERERERM8CJkNERERERGSVmAwREREREZFVYjJERERERERWicmQhUVHR0OlUulMvr6+OvOrV68OJycnuLm5oVWrVvjll1/01nPkyBG8+uqr8PHxgUajQdWqVTFw4ED88ccfRYprwYIFCAoKgkajQUhICPbt26fM69evn17ML7zwgsVjKqxt8s/TTrNnz1bqBAYGQqVSYe3atXrrr1mzJlQqFZYtW2Zyu+zduxcdOnSAn58fVCoVNm7cqDPfnJhUKhVsbGzg5+eHAQMG4L///jM5jrymT5+O5557Di4uLvD29kbnzp1x+vRpnToigujoaPj5+cHBwQHNmjXDyZMnderkjUs7lS9fvkgx5Y1NpVJh+PDhOuWpqano2LEjXF1d4eLighdeeEFnVMnAwEDExMQ80rYtcWzlbRMHBwcEBgaie/fu2L17d5FiKqz/mHpsAcC0adNgY2ODGTNmFCmWJ52hvmPK8aVlyfZ51POhoWNLpVKZFZslz4fa/ly9enXMnj0bRR0X6eLFi+jVqxc8PDzg6OiIevXq4dChQ8p8U847wMP3ih49eqBs2bJQq9UICAhA+/btsXnz5iLHVtT+Y8nzc2Hn5vv37+ODDz5A7dq14eTkBD8/P/Tp0weXLl3SWY8l+s/ChQtRp04d5UcwGzVqhB9//FGZb24/tsR7V16W6EuWeN+wxGcfwLLnH0u9do/aNlqFvY/lNWjQIKhUKr1tWzKeksBkqBjUrFkTGRkZynTixAllXtWqVfH555/jxIkT+PnnnxEYGIjw8HD8888/Sp0tW7bghRdeQHZ2NlavXo3U1FSsXLkSrq6umDBhgtnxxMXFYfjw4Rg/fjyOHDmCxo0bo23btjofVtu0aaMT87Zt23TWYamYjLVN3vKMjAwsWbIEKpVKZxh2APD398fSpUt1yg4cOIDMzEw4OTmZ0zS4desW6tati88//9zgfFNj0g41n56ejtWrV2Pv3r0YNmyYWbFoJSYm4u2338aBAwcQHx+PBw8eIDw8HLdu3VLqzJo1C/PmzcPnn3+OlJQU+Pr6onXr1rhx44bBuLTTkSNHihQTAKSkpGDRokWoU6eOTvmff/6Jl156CdWrV0dCQgKOHTuGCRMmQKPRFHlbBXnUYwv4vzY5ffo0VqxYgTJlyqBVq1aYOnWq2fEU1n+Awo8traVLl+L999/HkiVLzI4DsMwbLAAkJSWhXbt2cHNzg0ajQe3atTF37lzk5OQUKS6g4L5j6vEFPHr7aFnifAjoH1sZGRkYOnSoWbFY4nyojSM1NRWjR4/GuHHjsGjRIjNbBfjvv//w4osvws7ODj/++CNOnTqFuXPnokyZMkodU847P/zwA1544QXcvHkTy5cvx6lTp7Bu3Tp07twZH374Ia5fv252bI/afyx1fi7s3Hz79m0cPnwYEyZMwOHDh7Fhwwb88ccf6Nixo966HrX/lC9fHjNmzMDBgwdx8OBBtGjRAp06ddJJKMzpx5Z479KyVF96VJY61gHLnX8Ay712lmLK+xgAbNy4Eb/88gv8/PyKLZYSI2RREydOlLp165pc//r16wJAdu7cKSIit27dEk9PT+ncubPB+v/995/ZMT3//PMyePBgnbLq1avLmDFjRESkb9++0qlTpwKXt1RM5rZNp06dpEWLFjplAQEBMmbMGFGr1ZKenq6UDxw4UIYOHSqurq6ydOlSk7eRFwD5/vvvixTTJ598olM2efJkqVGjRpHiyO/y5csCQBITE0VEJDc3V3x9fWXGjBlKnbt374qrq6vExsYajauobty4IVWqVJH4+Hhp2rSpvPvuu8q8Hj16SK9evYwub4lYHvXYMhbHRx99JKVKlZLff/+9yPEZ6j+FHVtaCQkJUq5cObl37574+fkpr7U5Nm3aJFu3bpXTp0/L6dOnZdy4cWJnZye//fabEkubNm0kIyNDma5evaqzjg0bNoitra0MHDhQjhw5ImlpafLVV1+Jm5ubdOvWTXJzc82Oy1jfyc/Q8SVimfbRetTzoUjJ9GdTzz0NGjSQrl27mh3PBx98IC+99FKB800579y8eVM8PDykS5cuRtdjjkftP8V5fs5/bjbk119/FQBy/vx5ozFZgpubm3z99dciUvR+bIm2sURfKig+c1jiWBex7PmnIJZ47SyhoM9Bf/31l5QrV05+++03g9surngeF14ZKgZnzpyBn58fgoKC8Nprr+F///ufwXr37t3DokWL4Orqirp16wIAfvrpJ1y5cgXvv/++wWXyfrNiinv37uHQoUMIDw/XKQ8PD0dSUpLyf0JCAry9vZVb3y5fvqzMs2RMprbN33//ja1bt2LAgAF683x8fBAREYHly5cDePhtXFxcHPr3729yHEVhLKa8Ll68iC1btljsR4O136S6u7sDANLS0pCZmanzmqrVajRt2lTnNbWkt99+Gy+//DJatWqlU56bm4utW7eiatWqiIiIgLe3Nxo2bGj0MvujeJRjy5h3330XIoIffvjB0iEbPba0Fi9ejNdffx12dnZ4/fXXsXjxYrO306FDB7Rr1w5Vq1ZF1apVMXXqVDg7O+PAgQNKHbVaDV9fX2XS9ing4beDAwcORMeOHbFo0SLUq1cPgYGBiIyMxPLly/Hdd9/h22+/NTuugvpOfsaOL0u0D2CZ86ElWeJ8qCUiSEhIQGpqKuzs7MyOZdOmTQgNDcWrr74Kb29v1K9fH1999ZUy35Tzzo4dO3D16tUC3yuAh7e2mcMS/ScvS56f85+bC6qjUqnMfu82R05ODtauXYtbt26hUaNGSrm5/dhSbWOJvvSoLHmsW+r8Y4ilXrvilJubi969e+O9995DzZo1SyyOYlXS2dizZtu2bfLdd9/J8ePHlW+yfHx85MqVK0qdzZs3i5OTk6hUKvHz85Nff/1VmTdz5kwBIP/++69F4rl48aIAkP379+uUT506VapWrSoiImvXrpUtW7bIiRMnZNOmTVK3bl2pWbOm3L1716IxmdI2WjNnzhQ3Nze5c+eOTrn224eNGzdKpUqVJDc3V5YvXy7169cXESnWK0PGYrK3txcnJyfRaDQCQBo2bFikq3j55ebmSocOHXS+Zdu/f78AkIsXL+rUHThwoISHhxuMSzvNnz/f7BjWrFkjtWrVUvY777ezGRkZAkAcHR1l3rx5cuTIEZk+fbqoVCpJSEjQieVRvzV61GOrsDh8fHxkyJAhRY7PUP8p7NgSeXgFy9HRUY4ePSoiIkeOHBFHR0e5fv16kWN58OCBrFmzRuzt7eXkyZMi8vDbRldXV/Hy8pIqVapIZGSk/P3338oyGzZsEACSlJRkcJ1Vq1Y16VvUvIz1nfwKOr4s2T6WOB+KGD62nJycZM+ePSbHYqnzoTYOOzs7ASAajUZv/0yhVqtFrVbL2LFj5fDhwxIbGysajUaWL18uIqadd2bMmKH3XvHrr7/qtNHmzZtNjskS/ae4zs+Gzs353blzR0JCQqRnz54FxlTU/iMicvz4cXFychIbGxtxdXWVrVu3KvPM7ceWbBtL9CVtfEV937DUsV4c52cRy7x2j+vK0LRp06R169bKVd1n8coQk6FidvPmTfHx8ZG5c+fqlJ05c0aSk5Olf//+EhgYqHwoMfRm8ii0J4T8H3CmTJki1apVM7jMpUuXxM7OTtavX18sMWkZahutatWqyTvvvKNXrj3g7t+/Lz4+PpKQkCBNmzaVzz77TESKNxkyFtP48ePlzJkz8scff8iuXbukYcOGEhYWJg8ePChSLFpRUVESEBAgFy5cUMq0bySXLl3SqRsZGSkREREG49JO5r7Jpaeni7e3t/JGIKL7gUTbv15//XWd5Tp06CCvvfaaTiyWPlGae2wVFoe3t7dERUUVOZ7C+o+I/rElIrJw4UKpVauWTr1atWrJl19+aXYMj/IGqz3OC+ojHTt2lODgYJNjKazv5FfQ8WXJ9rHE+VDE8LF15swZuX37ttkxaRX1fKiNIykpSZo3by4ff/xxkbZvZ2cnjRo10ikbOnSovPDCCyJi2nnH0HvF3bt3lfYx5RjRslT/Ka7zs6Fzc1737t2TTp06Sf369fU+OFuq/2RnZ8uZM2ckJSVFxowZI56ensqXH/kV1o8t2TaW6Eva+B41GXrUY92S55+8LPHaPY5k6ODBg+Lj46OTuD6LyRBvkytmTk5OqF27Ns6cOaNTVrlyZbzwwgtYvHgxbG1tlcuuVatWBQD8/vvvFtm+p6cnbGxskJmZqVN++fJl+Pj4GFymbNmyCAgIUGK2dExahtoGAPbt24fTp08jMjKywGVtbW3Ru3dvTJw4Eb/88gt69uxp0djyKywmT09PVK5cGVWqVEGLFi0QExODpKQk7Nmzp8jbHDp0KDZt2oQ9e/bojAKnHXHKlNdUG5d2MvdWjUOHDuHy5csICQmBra0tbG1tkZiYiE8//RS2trbw8PCAra0tatSoobNccHCwzkOqxcHcY8uYq1ev4p9//kFQUFBxhqx3bAHAkiVLcPLkSaV9bW1tcfLkySLdilGtWjUcPXoUBw4cwJAhQ9C3b1+cOnUKANCjRw+8/PLLqFWrFjp06IAff/wRf/zxB7Zu3aqzDilgtC8Rgb29vcmxFNZ38g7IYOz4smT7WOJ8mHddeY+typUrw8HBweyYtIp6PtTG0ahRI6xfvx6ffPIJdu7cafb2y5Yta/Q4NuW8U6VKFQDQGWFNrVYr7WMOS/UfwPLn54LOzVr3799H9+7dkZaWhvj4eJQuXbrAmB6l/9jb26Ny5coIDQ3F9OnTUbduXcyfP99g3cL6sSXfuyzRlx6VpY51S55/8rLEa/c47Nu3D5cvX0aFChWU/T9//jxGjRqFwMDAxx5PcWEyVMyys7ORmpqKsmXLFlhHRJCdnQ3g4f2snp6emDVrlsG6165dM2v79vb2CAkJQXx8vE55fHw8wsLCDC5z9epVXLhwQYnZ0jFpFdQ2ixcvRkhISKHPevTv3x+JiYno1KkT3NzcihSDqUyNScvGxgYAcOfOHbO3JSJ45513sGHDBuzevVvvA3pQUBB8fX11XtN79+4hMTGxwNe0qFq2bIkTJ07g6NGjyhQaGoqePXvi6NGjUKvVeO655/SG/v7jjz8QEBBg0VjyM/fYMmb+/PkoVaoUOnfubMEI9eU/tk6cOIGDBw8iISFBp4337t2LlJQU/Pbbb2at/1HeYLUfZFNTUw3W//3335UvRkxRWN/RHiNAwcdXcbTPo54Pi8ujng8BwM3NDUOHDsXo0aPNHsL6xRdfNHocm3LeCQ8Ph7u7O2bOnGnWtg2xRP8pSFHPz4Wdm4H/S4TOnDmDnTt3wsPDw6xtPApj5ztT+/GjvHdpWaIvPSpLHOuWPv8YY4nXrjj07t0bx48f19l/Pz8/vPfee/jpp58eezzFpsSuST2jRo0aJQkJCfK///1PDhw4IO3btxcXFxc5d+6c3Lx5U8aOHSvJycly7tw5OXTokAwYMEDUarUy4pOIyMaNG8XOzk46dOgg8fHxkpaWJikpKfLee+9Jjx49zI5p7dq1YmdnJ4sXL5ZTp07J8OHDxcnJSc6dOyc3btyQUaNGSVJSkqSlpcmePXukUaNGUq5cOcnKyrJoTMbaRkt7f+7ChQsNriP/pdgrV67o3Fpg7m1yN27ckCNHjsiRI0cEgPLcS96Rf0yJafLkyZKRkSGXLl2SX375RZo2bSqenp4G7/8vzJAhQ8TV1VUSEhJ0Rv7Ku58zZswQV1dX2bBhg5w4cUJef/11KVu2rM5rVlyXrfPfqrJhwwaxs7OTRYsWyZkzZ+Szzz4TGxsb2bdvn04so0ePVtpaO+UfzcwYSxxbeV+r9PR0SUxMlIEDB4pKpdIZ2chUxvqPKcfWu+++Kw0bNjS47rCwMBk+fLjZMeXVokUL6du3r8F5V65cEbVardzHf/PmTXF3dzc4EtkPP/wgAHRuuysKQ7c5GTu+iqN9LHE+zNuP8k7mPEdQHOdDkYcjnGk0Glm3bp1Z7fLrr7+Kra2tTJ06Vc6cOSOrV68WR0dHWbVqlVLHlPOO9nzQrl072b59u/z5559y7Ngx5dnTTZs2mRVXXub2HxHLnp8LOzffv39fOnbsKOXLl5ejR4/q1MnOzjYYU1H7z9ixY2Xv3r2SlpYmx48fl3HjxkmpUqVkx44dRerHlnjv0rJUX3rU97BHPdaL6/xsqdfOkqPFFvY5KK9n8TY5JkMW1qNHDylbtqzY2dmJn5+fdO3aVbkP9M6dO9KlSxfx8/MTe3t7KVu2rHTs2FHvIW8RkZSUFOnatat4eXmJWq2WypUry1tvvSVnzpwpUlxffPGF8rBkgwYNlKEhb9++LeHh4eLl5SV2dnZSoUIF6du3r86w1ZaKyVjbaH355Zfi4OAg165dM7iOwg44c5OhPXv2CAC9Ke8HSFNiyrusl5eXtGvXTo4cOWJyHHkZigeAzn7l5ubKxIkTxdfXV9RqtTRp0kROnDihF9fjSIZERBYvXiyVK1cWjUYjdevWlY0bN+rFUtg+FcYSx1beOOzt7aVChQrSvXt32b17d5Hawlj/KezYys7OFg8PD5k1a5bBdc+dO1c8PT11PkQZY4k32HXr1omNjY0MHDhQjh07JmlpafL111+Lm5ubREZGFqmN8jLUdwo6vizdPnk96vmwoP48aNAgk2MozvPhwIEDpWbNmpKTk2NyPCIPByCpVauWqNVqqV69uixatEhnvinnHZGH7xXdunUTb29vsbW1FQ8PD4mIiJC1a9cWaXh2LXP6j5Ylz8+FnZvT0tIKrJN3cARL9J/+/fsrfdjLy0tatmwpO3bsEJGi9+NHfe/KyxJ9yd/fX3keuKiKeqwX5/nHEq+dJdpGy5TPQXkZOu9YMp6SoBIp4s9BExHRE2PAgAHYtWsXMjIy4Orqijp16uCDDz5A69atcefOHXTu3BlHjhzBtWvXULZsWTRv3hwff/wx/P39ddazb98+TJ06FcnJycjKygIAzJgxAx988EFJ7BYRWaGcnByULl0ay5cvR7du3Uo6nCfKk9Y2T1o8RcFkiIiIDLp79y46deqECxcuIDExEV5eXiUdEhE94/766y+sWLECkydPxrlz55QBF+jJa5snLZ6isi3pAIiI6Mmk0Wjwww8/ICYmBnv37sUrr7xS0iER0TOuXr168PDwwMqVK5/aD9fF5UlrmyctnqLilSEiIiIiIrJKHFqbiIiIiIisEpMhIiIiIiKySkyGiIiIiIjIKjEZIiIiIiIiq8RkiIiIiIiIrBKTISIiIiIiskpMhoiIiIiIyCoxGSIiIiIiIqvEZIiIiIiIiKzS/wNvm9dn0w2GuwAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1000x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "top_index=np.argsort(-np.asarray(data.iloc[:,1:].sum()))[:20]\n",
    "top_MSC=data.columns[1:][top_index] \n",
    "\n",
    "ind=np.arange(20)\n",
    "width=0.2\n",
    "\n",
    "fig=plt.figure(figsize=(10,4))\n",
    "ax=fig.add_subplot(111)\n",
    "ax.bar(x=ind, height=y_train[top_MSC].sum(axis=0)/y_train.shape[0], width=width, color='r')\n",
    "ax.bar(x=ind+width, height=y_val[top_MSC].sum(axis=0)/y_val.shape[0], width=width, color='b')\n",
    "ax.bar(x=ind+2*width, height=y_test[top_MSC].sum(axis=0)/y_test.shape[0], width=width, color='g')\n",
    "ax.set_xticks(ind+width)\n",
    "ax.set_xticklabels(top_MSC)\n",
    "\n",
    "ax.set_title(\"Frequencies of top twenty labels normalized by the dataset's size\")\n",
    "ax.legend(('Training set','Validation set','Test set'))\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3536835d",
   "metadata": {},
   "source": [
    "Tokenizing the data based on the new training set (with `maxlen=100` , `padding='post'` as before):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6f2a18fc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary size (based on the training set): 68006\n"
     ]
    }
   ],
   "source": [
    "tokenizer=Tokenizer()\n",
    "tokenizer.fit_on_texts(X_train)\n",
    "\n",
    "X_train=tokenizer.texts_to_sequences(X_train)\n",
    "X_train=pad_sequences(X_train, maxlen=maxlen, padding='post')\n",
    "\n",
    "X_val=tokenizer.texts_to_sequences(X_val)\n",
    "X_val=pad_sequences(X_val, maxlen=maxlen, padding='post')\n",
    "\n",
    "X_test=tokenizer.texts_to_sequences(X_test)\n",
    "X_test=pad_sequences(X_test, maxlen=maxlen, padding='post')\n",
    "\n",
    "# Saving the number of tokens \n",
    "vocabulary_size=len(tokenizer.word_counts)\n",
    "print(f'Vocabulary size (based on the training set): {vocabulary_size}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "222e3b9c",
   "metadata": {},
   "source": [
    "Since the MSC prediction problem is more complex, we increase the embedding dimesnion, the number of attention heads, and their complexity. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "6441e43f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional_1\"</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1mModel: \"functional_1\"\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ input_layer (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ token_and_position_embedding    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)        │     <span style=\"color: #00af00; text-decoration-color: #00af00\">4,358,848</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">TokenAndPositionEmbedding</span>)     │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)        │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ transformer_block               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">100</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)        │        <span style=\"color: #00af00; text-decoration-color: #00af00\">83,200</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">TransformerBlock</span>)              │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling1d        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)             │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">GlobalAveragePooling1D</span>)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_4 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)             │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_2 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">200</span>)            │        <span style=\"color: #00af00; text-decoration-color: #00af00\">13,000</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_5 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">200</span>)            │             <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">531</span>)            │       <span style=\"color: #00af00; text-decoration-color: #00af00\">106,731</span> │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                   \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ input_layer (\u001b[38;5;33mInputLayer\u001b[0m)        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m)            │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ token_and_position_embedding    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m64\u001b[0m)        │     \u001b[38;5;34m4,358,848\u001b[0m │\n",
       "│ (\u001b[38;5;33mTokenAndPositionEmbedding\u001b[0m)     │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (\u001b[38;5;33mDropout\u001b[0m)               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m64\u001b[0m)        │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ transformer_block               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m64\u001b[0m)        │        \u001b[38;5;34m83,200\u001b[0m │\n",
       "│ (\u001b[38;5;33mTransformerBlock\u001b[0m)              │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_average_pooling1d        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m)             │             \u001b[38;5;34m0\u001b[0m │\n",
       "│ (\u001b[38;5;33mGlobalAveragePooling1D\u001b[0m)        │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_4 (\u001b[38;5;33mDropout\u001b[0m)             │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m)             │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_2 (\u001b[38;5;33mDense\u001b[0m)                 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m200\u001b[0m)            │        \u001b[38;5;34m13,000\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_5 (\u001b[38;5;33mDropout\u001b[0m)             │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m200\u001b[0m)            │             \u001b[38;5;34m0\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_3 (\u001b[38;5;33mDense\u001b[0m)                 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m531\u001b[0m)            │       \u001b[38;5;34m106,731\u001b[0m │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">4,561,779</span> (17.40 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m4,561,779\u001b[0m (17.40 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">4,561,779</span> (17.40 MB)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m4,561,779\u001b[0m (17.40 MB)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "embed_dim = 64            # Embedding size for each token\n",
    "num_heads = 4             # Number of attention heads\n",
    "ff_dim = 128              # Hidden layer size in feed forward network inside transformer\n",
    "\n",
    "inputs = layers.Input(shape=(maxlen,))\n",
    "embedding_layer = TokenAndPositionEmbedding(maxlen=maxlen, input_dim=vocabulary_size+1, embed_dim=embed_dim)\n",
    "x = embedding_layer(inputs)\n",
    "x = layers.Dropout(0.1)(x)                                           \n",
    "transformer_block = TransformerBlock(embed_dim, num_heads, ff_dim,rate=0.2)\n",
    "x = transformer_block(x)\n",
    "x = layers.GlobalAveragePooling1D()(x)\n",
    "x = layers.Dropout(0.2)(x)\n",
    "x = layers.Dense(200, activation=\"relu\")(x)\n",
    "x = layers.Dropout(0.1)(x)\n",
    "outputs = layers.Dense(531, activation=\"sigmoid\")(x)\n",
    "\n",
    "model = keras.Model(inputs=inputs, outputs=outputs)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f5bcd9c",
   "metadata": {},
   "source": [
    "The task is multi-label, so we should use `sigmoid` activation in the last layer, and  `loss='binary_crossentropy'` in training (see [here](https://stackoverflow.com/questions/59336899/which-loss-function-and-metrics-to-use-for-multi-label-classification-with-very/59433454#59433454)). The validation set in hand will be used to evaluate loss for early stopping. Just like above, we set `patience=2` , `restore_best_weights=True`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "03497e95",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m405s\u001b[0m 120ms/step - auc: 0.4907 - loss: 0.0995 - val_auc: 0.5023 - val_loss: 0.0256\n",
      "Epoch 2/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m383s\u001b[0m 117ms/step - auc: 0.5258 - loss: 0.0255 - val_auc: 0.6434 - val_loss: 0.0215\n",
      "Epoch 3/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m383s\u001b[0m 117ms/step - auc: 0.6753 - loss: 0.0211 - val_auc: 0.7437 - val_loss: 0.0180\n",
      "Epoch 4/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m369s\u001b[0m 113ms/step - auc: 0.7555 - loss: 0.0182 - val_auc: 0.7933 - val_loss: 0.0163\n",
      "Epoch 5/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m301s\u001b[0m 92ms/step - auc: 0.7963 - loss: 0.0166 - val_auc: 0.8084 - val_loss: 0.0156\n",
      "Epoch 6/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m284s\u001b[0m 87ms/step - auc: 0.8212 - loss: 0.0156 - val_auc: 0.8186 - val_loss: 0.0152\n",
      "Epoch 7/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m286s\u001b[0m 87ms/step - auc: 0.8359 - loss: 0.0148 - val_auc: 0.8231 - val_loss: 0.0149\n",
      "Epoch 8/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m290s\u001b[0m 89ms/step - auc: 0.8485 - loss: 0.0143 - val_auc: 0.8258 - val_loss: 0.0148\n",
      "Epoch 9/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m288s\u001b[0m 88ms/step - auc: 0.8606 - loss: 0.0137 - val_auc: 0.8302 - val_loss: 0.0148\n",
      "Epoch 10/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m263s\u001b[0m 80ms/step - auc: 0.8705 - loss: 0.0132 - val_auc: 0.8335 - val_loss: 0.0147\n",
      "Epoch 11/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m249s\u001b[0m 76ms/step - auc: 0.8824 - loss: 0.0128 - val_auc: 0.8274 - val_loss: 0.0148\n",
      "Epoch 12/20\n",
      "\u001b[1m3272/3272\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m253s\u001b[0m 77ms/step - auc: 0.8862 - loss: 0.0125 - val_auc: 0.8281 - val_loss: 0.0149\n",
      "Epoch 12: early stopping\n",
      "Restoring model weights from the end of the best epoch: 10.\n"
     ]
    }
   ],
   "source": [
    "es=EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=2, restore_best_weights=True)\n",
    "\n",
    "opt=Adam(learning_rate=0.0001)\n",
    "# Keras AUC metric: https://keras.io/api/metrics/classification_metrics/\n",
    "model.compile(optimizer=opt, loss=\"binary_crossentropy\", metrics=[keras.metrics.AUC(multi_label=True,num_labels=531)])\n",
    "history = model.fit(X_train, (y_train), \n",
    "                    batch_size=32, epochs=20, validation_data=(X_val, y_val),\n",
    "                    callbacks=[es])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "23860eee",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "def plot_result(item):\n",
    "    plt.plot(history.history[item], label=item)\n",
    "    plt.plot(history.history[\"val_\" + item], label=\"val_\" + item)\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.locator_params(axis='x', integer=True)\n",
    "    plt.ylabel(item)\n",
    "    plt.title(\"Train and Validation {} Over Epochs\".format(item), fontsize=14)\n",
    "    plt.legend()\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "plot_result(\"loss\")\n",
    "plot_result(\"auc\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3dcd66f0",
   "metadata": {},
   "source": [
    "The trained model outputs a probability for each of the 531 MSC classes. Since the problem is multi-label, they do not necessarily add up to 1, and for each of them a threshold must be picked. We use the validation set for thresholding. First, let's try a uniform threshold across labels: we pick a threshold that maximizes the accuracy over the validation set. For that threshold, over the test set, various metrics will be computed along with the $F_1$ scores for the binary classification problems corresponding to the most frequent labels. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "e690faea",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m580/580\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 18ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m1282/1282\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m25s\u001b[0m 20ms/step\n",
      "\n",
      "Metrics for predicting 3-character MSC classes based on the cleaned text with the uniform thereshold 0.47000000000000003:\n",
      "\n",
      " \n",
      " Test Accuracy:  13.05%\n",
      " Test F1-score micro: 35.77%\n",
      " Test F1-score macro: 13.47%\n",
      " Test F1-score weighted: 31.74%\n",
      " Test average Jaccard score: 28.44%\n",
      "\n",
      "Test F1-scores for the top twenty most frequent labels:\n",
      " [63.44, 67.92, 62.35, 64.22, 54.38, 62.38, 62.54, 42.28, 51.15, 44.53, 41.05, 38.25, 43.4, 39.78, 37.08, 31.16, 36.02, 47.89, 33.3, 50.22]\n"
     ]
    }
   ],
   "source": [
    "y_pred_val_proba=model.predict(X_val)\n",
    "thresholds=np.linspace(0,1,101)\n",
    "scores=np.zeros(101)\n",
    "\n",
    "for i in range(101):\n",
    "    scores[i]=100*accuracy_score(y_val, (y_pred_val_proba>thresholds[i]).astype(int))\n",
    "    \n",
    "plt.plot(thresholds,scores)\n",
    "plt.xlabel('threshold')\n",
    "plt.ylabel('Accuracy score')\n",
    "plt.title('The accuracy score over the validation set for different probability thresholds', fontsize=9)\n",
    "plt.show()\n",
    "\n",
    "threshold=thresholds[np.argmax(scores)]\n",
    "y_pred_proba=model.predict(X_test)\n",
    "y_pred=(y_pred_proba>threshold).astype(int)\n",
    "\n",
    "print(f'\\nMetrics for predicting 3-character MSC classes based on the cleaned text with the uniform thereshold {threshold}:\\n')\n",
    "print(f' \\n Test Accuracy:  {round(100*accuracy_score(y_test, y_pred, normalize=True),2)}%')\n",
    "print(f\" Test F1-score micro: {round(100*f1_score(y_test, y_pred, average='micro'),2)}%\")\n",
    "print(f\" Test F1-score macro: {round(100*f1_score(y_test, y_pred, average='macro'),2)}%\")\n",
    "print(f\" Test F1-score weighted: {round(100*f1_score(y_test, y_pred, average='weighted'),2)}%\") \n",
    "print(f\" Test average Jaccard score: {round(100*jaccard_score(y_test, y_pred, average='samples'),2)}%\") \n",
    "\n",
    "list_top_f1=[]\n",
    "MSC_list=y_val.columns.to_list()\n",
    "for top_class in top_MSC:\n",
    "    column_index=MSC_list.index(top_class)\n",
    "    list_top_f1+=[round(100*f1_score(y_test[top_class],y_pred[:,column_index]),2)]\n",
    "    \n",
    "print(f'\\nTest F1-scores for the top twenty most frequent labels:\\n {list_top_f1}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7da8141c",
   "metadata": {},
   "source": [
    "It is also possible to choose thresholds separately. Meaning, for each label, we pick a threshold that maximizes the accuracy of the corresponding binary classifier. This can be done in a highly-vectorized fashion. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5a32f947",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Different probability thresholds chosen for different labels:\n",
      " [0.63 0.06 0.04 0.4  0.11 0.06 0.34 0.46 0.39 0.77 0.52 0.12 0.06 0.56\n",
      " 0.1  0.41 0.56 0.41 0.06 0.52 0.12 0.42 0.39 0.56 0.66 0.37 0.03 0.41\n",
      " 0.63 0.58 0.64 0.46 0.58 0.48 0.48 0.03 0.01 0.15 0.56 0.09 0.43 0.53\n",
      " 0.55 0.34 0.24 0.01 0.05 0.56 0.48 0.16 0.46 0.08 0.38 0.46 0.51 0.27\n",
      " 0.61 0.01 0.14 0.35 0.59 0.05 0.02 0.12 0.1  0.51 0.63 0.33 0.61 0.11\n",
      " 0.13 0.46 0.02 0.34 0.2  0.36 0.47 0.03 0.01 0.48 0.17 0.02 0.25 0.66\n",
      " 0.28 0.14 0.68 0.47 0.11 0.2  0.34 0.4  0.34 0.54 0.39 0.14 0.14 0.62\n",
      " 0.54 0.64 0.11 0.37 0.01 0.68 0.5  0.45 0.04 0.45 0.21 0.28 0.2  0.52\n",
      " 0.5  0.45 0.4  0.12 0.36 0.55 0.03 0.24 0.21 0.05 0.56 0.51 0.31 0.58\n",
      " 0.03 0.38 0.14 0.4  0.51 0.65 0.16 0.58 0.61 0.46 0.58 0.29 0.16 0.64\n",
      " 0.09 0.44 0.16 0.37 0.43 0.33 0.07 0.52 0.17 0.08 0.09 0.56 0.4  0.53\n",
      " 0.4  0.72 0.57 0.05 0.03 0.44 0.48 0.15 0.74 0.48 0.1  0.06 0.14 0.01\n",
      " 0.18 0.03 0.63 0.02 0.54 0.1  0.33 0.02 0.27 0.52 0.08 0.06 0.48 0.53\n",
      " 0.09 0.13 0.19 0.69 0.02 0.51 0.01 0.01 0.5  0.12 0.12 0.1  0.55 0.26\n",
      " 0.11 0.34 0.22 0.45 0.17 0.04 0.02 0.3  0.48 0.49 0.3  0.31 0.29 0.69\n",
      " 0.3  0.54 0.25 0.29 0.24 0.32 0.33 0.24 0.18 0.63 0.02 0.34 0.28 0.37\n",
      " 0.21 0.06 0.48 0.65 0.53 0.68 0.37 0.29 0.55 0.31 0.19 0.5  0.7  0.34\n",
      " 0.06 0.59 0.11 0.24 0.04 0.08 0.51 0.29 0.44 0.62 0.02 0.28 0.05 0.42\n",
      " 0.27 0.03 0.56 0.6  0.4  0.46 0.43 0.54 0.22 0.15 0.03 0.11 0.36 0.63\n",
      " 0.54 0.34 0.7  0.59 0.4  0.57 0.12 0.34 0.46 0.45 0.04 0.2  0.52 0.4\n",
      " 0.28 0.05 0.55 0.04 0.01 0.43 0.39 0.08 0.74 0.56 0.45 0.15 0.38 0.56\n",
      " 0.38 0.62 0.02 0.34 0.01 0.35 0.52 0.25 0.53 0.11 0.49 0.6  0.03 0.56\n",
      " 0.4  0.04 0.16 0.44 0.35 0.61 0.39 0.05 0.05 0.34 0.28 0.64 0.6  0.49\n",
      " 0.22 0.48 0.63 0.11 0.16 0.03 0.58 0.17 0.48 0.01 0.24 0.53 0.62 0.47\n",
      " 0.23 0.14 0.05 0.56 0.34 0.28 0.37 0.54 0.01 0.02 0.4  0.01 0.08 0.4\n",
      " 0.02 0.01 0.13 0.47 0.08 0.02 0.03 0.51 0.51 0.47 0.3  0.23 0.59 0.07\n",
      " 0.57 0.6  0.28 0.68 0.35 0.2  0.48 0.51 0.25 0.17 0.62 0.01 0.41 0.06\n",
      " 0.46 0.51 0.09 0.01 0.38 0.52 0.31 0.56 0.06 0.45 0.06 0.44 0.02 0.09\n",
      " 0.53 0.26 0.12 0.03 0.29 0.32 0.65 0.47 0.27 0.02 0.66 0.05 0.56 0.42\n",
      " 0.44 0.41 0.47 0.6  0.09 0.1  0.02 0.35 0.55 0.12 0.22 0.53 0.22 0.63\n",
      " 0.29 0.13 0.42 0.06 0.06 0.56 0.16 0.27 0.05 0.39 0.27 0.55 0.31 0.5\n",
      " 0.53 0.52 0.16 0.09 0.17 0.44 0.29 0.47 0.33 0.06 0.55 0.35 0.51 0.14\n",
      " 0.52 0.49 0.64 0.51 0.05 0.37 0.09 0.46 0.26 0.46 0.33 0.08 0.21 0.22\n",
      " 0.58 0.01 0.66 0.57 0.47 0.51 0.21 0.01 0.64 0.6  0.07 0.07 0.45 0.01\n",
      " 0.01 0.47 0.23 0.03 0.61 0.42 0.05 0.28 0.64 0.35 0.44 0.17 0.53 0.5\n",
      " 0.21 0.47 0.5  0.38 0.03 0.48 0.12 0.54 0.11 0.38 0.03 0.03 0.28 0.14\n",
      " 0.15 0.02 0.49 0.09 0.35 0.44 0.6  0.01 0.44 0.56 0.27 0.39 0.22 0.06\n",
      " 0.12 0.54 0.6  0.59 0.07 0.03 0.42 0.63 0.4  0.4  0.64 0.2  0.19]\n",
      "\n",
      "Metrics for predicting 3-character MSC classes based on the cleaned text with varying thresholds:\n",
      "\n",
      " \n",
      " Test Accuracy:  12.81%\n",
      " Test F1-score micro: 32.62%\n",
      " Test F1-score macro: 14.83%\n",
      " Test F1-score weighted: 29.2%\n",
      " Test average Jaccard score: 26.15%\n",
      "\n",
      "Test F1-scores for the top twenty most frequent labels:\n",
      " [61.17, 67.17, 63.38, 62.75, 54.05, 60.09, 57.28, 42.28, 51.28, 42.66, 33.66, 29.62, 38.88, 27.97, 26.59, 13.08, 16.3, 43.39, 29.58, 43.86]\n"
     ]
    }
   ],
   "source": [
    "def varying_thresholds(y_true,y_proba,n=100):\n",
    "    \"\"\"\n",
    "    For a multi-label classification problem, returns thresholds that maximize the accuracy of prediction for every label.  \n",
    "    \n",
    "    Input\n",
    "    -----\n",
    "    y_true: numpy array\n",
    "        The array of true labels, of shape data_size*(n_labels).\n",
    "    y_proba: numpy array\n",
    "        The array of predicted labels, of shape data_size*(n_labels).\n",
    "    n: int\n",
    "         Determines the thresholds to be examined, n+1 thresholds obtained from dividing [0,1] into n subintervals of length 1/n.\n",
    "         \n",
    "    Output\n",
    "    ------\n",
    "    A numpy array of thresholds, one for each label. \n",
    "    \"\"\"\n",
    "    assert isinstance(y_true,np.ndarray) and isinstance(y_proba,np.ndarray), \"[Error] The first two inputs should be numpy arrays.\"\n",
    "    assert y_true.ndim==2 and y_true.shape==y_proba.shape, \"[Error] The input arrays should be 2D and with the same shape.\"\n",
    "    assert isinstance(n,int) and n>0, \"[Error] The last (optional) input should a positive integer.\"\n",
    "     \n",
    "    thresholds=np.linspace(0,1,n+1)          \n",
    "    #A syntax for repeating an array a number of times along a new dimension:\n",
    "    thresholds_array=np.repeat(thresholds[None,:],y_true.shape[0],axis=0)        #size: data_size*(n+1)            \n",
    "    y_proba_duplicated=np.repeat(y_proba[:,None,:],n+1,axis=1)                   #size: data_size*(n+1)*(n_labels)\n",
    "    y_pred_labels=(y_proba_duplicated>thresholds_array[:,:,None]).astype(int)    #size: data_size*(n+1)*(n_labels)\n",
    "    y_true_duplicated=np.repeat(y_true[:,None,:],n+1,axis=1)                     #size: data_size*(n+1)*(n_labels)\n",
    "    accuracy_array=np.sum((y_pred_labels==y_true_duplicated).astype(int),axis=0) #size: (n+1)*(n_labels)\n",
    "    best_thresholds_index=np.argmax(accuracy_array,axis=0)                       #size: n_labels\n",
    "    return thresholds[best_thresholds_index]\n",
    "    \n",
    "    \n",
    "thresholds=varying_thresholds(np.asarray(y_val),y_pred_val_proba)\n",
    "print(f'Different probability thresholds chosen for different labels:\\n {thresholds}\\n')\n",
    "\n",
    "y_pred=(y_pred_proba>thresholds[None,:]).astype(int)\n",
    "\n",
    "print('Metrics for predicting 3-character MSC classes based on the cleaned text with varying thresholds:\\n')\n",
    "print(f' \\n Test Accuracy:  {round(100*accuracy_score(y_test, y_pred, normalize=True),2)}%')\n",
    "print(f\" Test F1-score micro: {round(100*f1_score(y_test, y_pred, average='micro'),2)}%\")\n",
    "print(f\" Test F1-score macro: {round(100*f1_score(y_test, y_pred, average='macro'),2)}%\")\n",
    "print(f\" Test F1-score weighted: {round(100*f1_score(y_test, y_pred, average='weighted'),2)}%\") \n",
    "print(f\" Test average Jaccard score: {round(100*jaccard_score(y_test, y_pred, average='samples'),2)}%\") \n",
    "\n",
    "list_top_f1=[]\n",
    "for top_class in top_MSC:\n",
    "    column_index=MSC_list.index(top_class)\n",
    "    list_top_f1+=[round(100*f1_score(y_test[top_class],y_pred[:,column_index]),2)]\n",
    "    \n",
    "print(f'\\nTest F1-scores for the top twenty most frequent labels:\\n {list_top_f1}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "165569d1",
   "metadata": {},
   "source": [
    "**Conclusion)** With either of the two thresholding conventions above, the performance over the test set of the transformer model is worse than that of the SVM model we had trained before (slightly better accuracy but other metrics are far worse). Therefore, on our data, SVM seems to work better than transformers for predicting Mathematical Subject Classes. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "97fee0d3",
   "metadata": {},
   "source": [
    "A useful link\n",
    "\n",
    "----------------------------------------------------------------------------------------\n",
    "\n",
    "[A curated list of learning resources on transformers and LLMs, carefully divided to different topics.](https://drive.google.com/file/d/13a74b1w1Ecb9ZHvPvrFpQfAH1U2dUERh/view?usp=sharing)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}