LGAIMLJan 3, 2021

StarNet: Gradient-free Training of Deep Generative Models using Determined System of Linear Equations

arXiv:2101.00574v1
Originality Highly original
AI Analysis

This work addresses the computational challenges of gradient-based training for deep generative models, offering a potentially more scalable and stable alternative for researchers and practitioners in machine learning.

This paper introduces StarNet, a deep generative model that can be trained by solving determined systems of linear equations, eliminating the need for gradient-based optimization. This approach offers high scalability for estimating latent codes and model parameters, and provides desirable least-square bounds for these estimations within each layer.

In this paper we present an approach for training deep generative models solely based on solving determined systems of linear equations. A network that uses this approach, called a StarNet, has the following desirable properties: 1) training requires no gradient as solution to the system of linear equations is not stochastic, 2) is highly scalable when solving the system of linear equations w.r.t the latent codes, and similarly for the parameters of the model, and 3) it gives desirable least-square bounds for the estimation of latent codes and network parameters within each layer.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes