Solving Regularized Exp, Cosh and Sinh Regression Problems

TL;DR

This paper introduces a convex regularized exponential regression framework inspired by attention mechanisms in large language models, and proposes an efficient approximate Newton method leveraging input sparsity for large-scale problems.

Contribution

It develops a novel convex regularized exponential regression model and an input sparsity time algorithm using approximate Newton methods, improving scalability for large datasets.

Findings

The proposed method achieves input sparsity time complexity per iteration.

It effectively solves regularized exponential regression problems with functions like exp, cosh, sinh.

The algorithm converges in a logarithmic number of iterations relative to the desired accuracy.

Abstract

In modern machine learning, attention computation is a fundamental task for training large language models such as Transformer, GPT-4 and ChatGPT. In this work, we study exponential regression problem which is inspired by the softmax/exp unit in the attention mechanism in large language models. The standard exponential regression is non-convex. We study the regularization version of exponential regression problem which is a convex problem. We use approximate newton method to solve in input sparsity time. Formally, in this problem, one is given matrix $A \in \mathbb{R}^{n \times d}$, $b \in \mathbb{R}^n$, $w \in \mathbb{R}^n$ and any of functions $\exp, \cosh$ and $\sinh$ denoted as $f$. The goal is to find the optimal $x$ that minimize $ 0.5 \| f(Ax) - b \|_2^2 + 0.5 \| \mathrm{diag}(w) A x \|_2^2$. The straightforward method is to use the naive Newton's method. Let $\mathrm{nnz}(A)$ denote the number of non-zeros entries in matrix $A$. Let $\omega$ denote the exponent of matrix multiplication. Currently, $\omega \approx 2.373$. Let $\epsilon$ denote the accuracy error. In this paper, we make use of the input sparsity and purpose an algorithm that use $\log ( \|x_0 - x^*\|_2 / \epsilon)$ iterations and $\widetilde{O}(\mathrm{nnz}(A) + d^{\omega} )$ per iteration time to solve the problem.