This is where I will be uploading Colab notebooks on work-in-progress projects. I only upload the latest version of Colab-notebooks. (Project names are in place because they sound cool and it is easy for me to remember these ones amongst the ocean of project ideas I have in my head!)
This is a LSTM network that will create sitar music. The initial architecture will be simple, but I plan to augment it by using the 'APOLLO' optimizer paper and using DeepMind's 'PowerPropagation' technique to boost performance, not to mention that I will also derive extra juice from my GPU via some CUDA programming.
This is a planned project that aims to use, implement and expand on DeepMind's research on neural cryptography by simulating a cryptographic system in which, out of the 3 neural networks in place, 2 will act as 'Bob' and 'Alice' (the guys who communicate via asymmetric key cryptographic methods) and the last network will act as the eavesdropper, attempting to crack their encryption. The paper I am inspired by is this one. Most probable architectures to be used will be RNNs and GANs. And of course, I will also use techniques from other papers to optimize these networks and experiment with my simulation!