I am Dhruvanshu Joshi, a Prefinal year student at VJTI, India majoring in computer science. I have been actively contributing to pymc for quite sometime now. I have previously worked under the mentorship of @ricardoV94 and @larryshamalama on the project aimed at implementing logprob derivation for order statistics through multiple prs. I would very much like to continue my work through the GSOC project Improve log-probability inference of order statistics.
Apart from that, I also loved the project idea Add PyTorch backend to PyTensor. This is my rough understanding on how to carry out that project: Identify the common functionalities of pytorch and pytensor and try to implement that in pymc wherein depending on the parameter passed, PyMC usses the pytorch backend or pytensor backend. Then we can try some complex operations also. I’d like to hear feedbacks on this idea and how to proceed further in order to create a comprehensive proposal.
I realise that I can work on only one idea and personally like both. I’d be happy if you could direct me towards any of these two projects.