AI Researcher

07/08 - Topological GNNs

• lympth nodes were swollen like little golf balls for the past 5 days -- had me concerned because the doctor said it was "really unusual with no cold symptoms" and another small lump started forming below my adams apple. Now, it's finally reducing exponentially after a nice 10hr sleep.
• Finished watching some topolgoical extensions of GNNs and actually conceptualise the power of them now. I can't wait to experiment with topolgoical-geometric modifications to GNNs!
  • Topological Machine Learning II: Topology-Based Graph Learning -- really interesting to see how GNNs are extended to handle topologies with being able to learn different filtrations. Looking deeper into topological GNNs and geometric DL
  • [Shapes, Spaces, Simplices, and Structure: Geometry, Topology, and Machine Learning
  • Binging a lot of Bastians videos both because they're really good and I have a new perspective with GNNs to make it more fascinating and understand deeply the concepts, and that I'm doing a podcast with him once again on the 12th. Also doing one with the legendary Mathilde Papillon ](https://www.youtube.com/watch?v=frjmKcwUsl8&ab_channel=BastianGrossenbacher-Rieck)
• Had an amazing week of Vidit Nandan and Micheal Bronstein follow me on twitter ! Both of them have so many interesting papers, especially Nandan with the many applications of Morse + Category Theory.
• Started reading a bunch of papers to prepare for podcasts. now think reading papers inspires ideas stronger than other forms of content bc it's a "public repo" of their math (sometimes applied) ideas. Also papers typically have debrief papers early on to prepare you for understanding the math -- often really concise and convey extremely well! Alongside papers, consuming books is the optimal way to quickly deeply understand and using LLMs to help reason about pitfalls.