corrected CUDA support + added optimizer in checkpoint + cleaning

Revert "updated broken link"

updated broken link

updated broken link

final clean + added folder structure

generated all figures + done some cleaning

modified debugging structure + found numerical instability example

updated README + added loading of Laplacians

cleaned some files + added mnist_012 model + new debug for mnist_rot

need to optimize update_mask func in DP layer

selection of datsets in command line + reorganization of repo

added report

updated loss and error + added freeze option for spectral conv weights

cleaned file

finished layer debug + added plot loss/error + done some param tuning

found another bug (matrix power)

continued debugging

continued debugging

more debugging

added theano source code file to save modifications

migrated code to 0.4.0 + began debugging

small changes

added loader for saved_dataset (from Theano) + reorganized few things

improved readibility of code + created new class for TIGraNet models

corrected bug + done some cleaning

changed nodes_of_interest to mask + done some cleaning

updated code following theano source code

corrected spectral conv layer + added debugging comments/functions

corrected bug from spectral conv layer + modified code to be more human…

modifications made from instance - new save name for model state (basically)

midterm presentation pdf

changed CUDA tensors

found bug in dataloader

added testing accuracy + function to load saved_model

added pretrained weigths (but validation not good)

corrected CUDA activations

updated .gitignore with the new folders

added function to save model state

corrected metrics calculation + activated shuffling when training

added new plot module

added new path module

corrected last layer + specified model mode for prediction

reorganized parameters

renamed variable

renamed argument + added comments

renamed argument + added CUDA option + implemented validation

added new parameters

renamed argument + added last layers

renamed argument

updated to run correctly

implemented visualization for the layers

implemented stat layer + rearranged few things

renamed functions + correctly implemented functions for stat layer

renamed variables according their meaning in all functions used so far

updated .gitignore

corrected DynamicPool layer (error + return V)

updated SpectralConv layer to take into account nodes of interest

updated .gitignore for personal files

new module containing models and custom layers

added some code to test training of model

added to_sparse func + some func needed later

updated laplacian function + deleted unused function

updated mean/std parameters of both dataset + added new graph module

updated imshow function depending on dataset

added parameter module

updated .gitignore

deleted unused module

transfer from github - completed loading datasets