This Repository contains the research I did relavent to CNNs. And below I will list down the important scripts and how to run them

##conv_theano CNN implemented in Theano (Obsolete)

##conv_tensorflow Tensorflow implementation of a CNN

conv_net_plot.py

• Contains a simple CNN. It will start a CNN of specified architecture and run it on cifar-10 dataset. The CNN architecture can be changed via setting hyparams,conv_ops,depth_conv parameters in the code

deconv_visualization.py

• Contains the code for visualizing high level convolutional filters using deconv paper

experiment_inc_initialization.py (Obsolete)

• An experimental script used to test if adding layers incrementally help

experiment_inc_initialization_v2.py

• Made several improvements over the v1 script. Instead of having conv_balance and pool_balance, got rid of conv_balance
• Can run this using python3 experiment_inc_initialization_v2.py --output_dir=<dir_to_save_all_logs_and_data>
• Use CUDA_VISIBLE_DEVICES=x,y to restrict GPU usage

##conv_tensorflow/adaptive_cnn