Hey guys! This is the summary of all the work I did for the umbrella organization CERN-HSF as a GSOC student.
The headers to layer classes can be found here. The DeepNet suited for AutoEncoders can be found here. All the tests for submodules can be found here. The VariableDAETransform class can be found here. All my commits can be found here and here. As we were three students working on a common repo to create a common DeepLearning module so the merge will be done when everything finishes off for everyone. For me whats left for now is to implement just 2 functions for CUDA Architecture. I will finish the CPU soon and will continue with CUDA after this program. Moreover for Reference Architecture everything is done.
Results after 100 epochs:
Graphs are plotted to analyse Reconstructed input and actual input. The x axis shows range of error and y axis shows number of times that error was observed. The results are for 4 variables that are there in Root’s dataset.
Result after 1500 epochs
So things improved a lot after 1500 epochs. :)