-
Complex-Analytic Proofs of Global Attraction for Neural Kernel Map
A global contraction of the kernel map using Schwarz–Pick, Julia–Carathéodory, and Rogosinski extremals.
-
From a Stochastic Traffic Jam to a Solvable PDE
An exploration of how complex, random traffic flow can be described by a simple, solvable continuous model.
-
NTK: A First Principles Derivation
A first principles derivation of the classic NTK result (no magic), with an analysis that already suggests the scaling for feature learning
-
The Mean-Field View of Deep Learning
A review of the most important papers in theory of signal propagation
-
On the global stability of activation variance
Can an activation have multiple variance stable points? This post proves it affirmitively.