Here are some basic tools I built on top of keras to facilitate deep learning experiments. The jupyter notebook EvaluationModels_Mnist2.ipynb runs these tools on the MNIST dataset as a proof of concept. The jupyter notebook can run on any image dataset with changes to a few lines of code.
- Python 3.6.8 with keras 2.1.6, hyperopt 0.1, and networkx 1.11.
- For Python 3.7 use the above versions of keras and networkx with hyperopt 0.1.2.
- In EvaulationModels_Mnist2.ipynb change local_dir to the location of your local directiory, especially if using google colab.
- Make sure to have the folder trained_models in the local directory for the code to work correctly.
A special thanks to Vik for giving me the idea to create this class for running deep learning experiments.