bulatuseinov / MIPT-OPT-2020

Quasi-Newton optimization methods for Deep Learning using PyTorch-Optimizer interface.

Quasi Newton optimization methods in context of Deep Learning

Research project for the course of optimization methods as MIPT.

Spring 2020

Authors:

• Alexander Zhogov
• Mikhail Sysak
• Bulat Useinov

Academic supervisor:

• Alexandr Katrutsa

In this article, we tried to demonstrate the suitability of Quasi Newton optimization methods for Deep Learning problems. We conducted a series of experiments and compared different approaches of optimization with each other.

(Only Russian version is currently available)