Roman Sarapin

collaborators

2 papers

cs.LG2026

Randomly initialized autoencoders: fixed points and edge-of-chaos

Leonid Berlyand, Roman Sarapin, Yitzchak Shmalo +2

In this paper we study autoencoders, a special class of deep neural nets (DNNs) whose performance can be characterized via their fixed points. This perspective naturally raises que…

math-ph2025

Limiting eigenvalue distribution of the general deformed Ginibre ensemble

Roman Sarapin

Consider the $n \times n$ matrix $X_{n} = A_{n} + H_{n}$ , where $A_{n}$ is a $n \times n$ matrix (either deterministic or random) and $H_{n}$ is a $n \times n$ matrix independent from $A_{n}$ drawn f…