Saved in:
Bibliographic Details
Main Authors: Jin, Yanhao, Balasubramanian, Krishnakumar, Lai, Lifeng
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.14183
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915144116731904
author Jin, Yanhao
Balasubramanian, Krishnakumar
Lai, Lifeng
author_facet Jin, Yanhao
Balasubramanian, Krishnakumar
Lai, Lifeng
contents We investigate the in-context learning capabilities of transformers for the $d$-dimensional mixture of linear regression model, providing theoretical insights into their existence, generalization bounds, and training dynamics. Specifically, we prove that there exists a transformer capable of achieving a prediction error of order $\mathcal{O}(\sqrt{d/n})$ with high probability, where $n$ represents the training prompt size in the high signal-to-noise ratio (SNR) regime. Moreover, we derive in-context excess risk bounds of order $\mathcal{O}(L/\sqrt{B})$ for the case of two mixtures, where $B$ denotes the number of training prompts, and $L$ represents the number of attention layers. The dependence of $L$ on the SNR is explicitly characterized, differing between low and high SNR settings. We further analyze the training dynamics of transformers with single linear self-attention layers, demonstrating that, with appropriately initialized parameters, gradient flow optimization over the population mean square loss converges to a global optimum. Extensive simulations suggest that transformers perform well on this task, potentially outperforming other baselines, such as the Expectation-Maximization algorithm.
format Preprint
id arxiv_https___arxiv_org_abs_2410_14183
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle In-context Learning for Mixture of Linear Regressions: Existence, Generalization and Training Dynamics
Jin, Yanhao
Balasubramanian, Krishnakumar
Lai, Lifeng
Machine Learning
We investigate the in-context learning capabilities of transformers for the $d$-dimensional mixture of linear regression model, providing theoretical insights into their existence, generalization bounds, and training dynamics. Specifically, we prove that there exists a transformer capable of achieving a prediction error of order $\mathcal{O}(\sqrt{d/n})$ with high probability, where $n$ represents the training prompt size in the high signal-to-noise ratio (SNR) regime. Moreover, we derive in-context excess risk bounds of order $\mathcal{O}(L/\sqrt{B})$ for the case of two mixtures, where $B$ denotes the number of training prompts, and $L$ represents the number of attention layers. The dependence of $L$ on the SNR is explicitly characterized, differing between low and high SNR settings. We further analyze the training dynamics of transformers with single linear self-attention layers, demonstrating that, with appropriately initialized parameters, gradient flow optimization over the population mean square loss converges to a global optimum. Extensive simulations suggest that transformers perform well on this task, potentially outperforming other baselines, such as the Expectation-Maximization algorithm.
title In-context Learning for Mixture of Linear Regressions: Existence, Generalization and Training Dynamics
topic Machine Learning
url https://arxiv.org/abs/2410.14183