[COMP6248 Reproductibility Challenge]

Are Neural Nets Modular? Inspecting the Functional Modularity of NNs using a differentiable weight mask

Paper: https://arxiv.org/abs/2010.02066

Original Code Base: https://github.com/RobertCsordas/modules/tree/efdb8790b074862581e035c9ab5bf889440a8023

This project includes :

• Data generation method (algorithmic data, +/* operations)
• Network pre-training (FNN) (model data can be found in networks/cache-networks/)
• Mask Generation (pre-generated masks can be found in trainedmasks/)
• Testing for Masked Network
• Calculating degree of weight sharing between two masks