Created: February 16, 2020
Modified: February 22, 2020
Modified: February 22, 2020
ongoing projects
This page is from my personal notes, and has not been specifically reviewed for public consumption. It might be incomplete, wrong, outdated, or stupid. Caveat lector.- These are things that I might plausibly decide I want to work on when I sit down on the weekend.
- Expanding nodes on this graph.
- Blogging: write up a post(s) for nonlinearfunction
- Monte Carlo Tree Search
- Understand DNA: continue exploring and writing up that doc.
- Step 2a once I feel confident about DNA: how do viruses reproduce?
- Step 2b: how does serotonin bind it its receptor sites, and how do drugs affect this?
- Graph neural nets:
- are there novel styles of normalizing flows?
- does the message-passing connection have more life to it?
- can we build generative models that are permutation-invariant? If not, why not?
- Probabilistic programming:
- Foundations: probabilistic programs are the universal models: the only model class that definitely contains your system. Read some of the early papers about this?
- Recent papers: things for a reading group?
- Structure learning: can we use continuous tools to learn discrete models?
- think about directions for probabilistic programming
- Understand relationship between backprop and message passing
- Value uncertainty: read papers on distributional RL. clarify the thesis that it feels like these approaches should be complementary.
- The mind: read papers on dopamine, psychedelics, predictive coding, etc.
- Causality: read Kevin's book chapter draft on causal inference. Implement a toy causal inference example: a world where I know the causal model, can generate synthetic data, and want to infer causal effects from the synthetic data. Think about how to encode this in a PPL.