Hi, I’m Henry. I’m a PhD Candidate in Pure Mathematics at Stanford, but I’m currently on leave to work at Etched.
Posts
-
Stategies for Learning and Maintaining Knowledge: A Proposal
In this post, I will muse about strategies for learning and maintaining knowledge. The hope is to find an optimal strategy, or at least an approximation. I would assume that standard methods of learning are inefficient. One thing I have found is that without reviewing material, I forget it quite quickly. I feel like I have the memory of a goldfish! However, if I choose to review everything I read at a fixed interval, the amount I have to review grows linearly in time and eventually becomes unmanageable.
-
Implicit Regularization and Generalization Of Deep Neural Networks
In this post I will reflect on what (little) I know about implicit generalization, and discuss future directions. Ultimately the goal is to show that the reason that deep networks generalize well, even without convergence, is that they are biased to learn simple functions, ultimately by their compute-limited nature.
subscribe via RSS