paper-with-me

홈 › Papers

High-Dimensional Learning Dynamics of Quantized Models with Straight-Through Estimator

2025-10-12 · Yuma Ichikawa, Shuhei Kashiwamura, Ayaka Sakata arxiv

Quantized neural network training optimizes a discrete, non-differentiable objective. The straight-through estimator (STE) enables backpropagation through surrogate gradients and is widely used. While previous studies have primarily focused on the properties of surrogate gradients and their convergence, the influence of quantization hyperparameters, such as bit width and quantization range, on learning dynamics remains largely unexplored. We theoretically show that in the high-dimensional limit, STE dynamics converge to a deterministic ordinary differential equation. This reveals that STE training exhibits a plateau followed by a sharp drop in generalization error, with plateau length depending on the quantization range. A fixed-point analysis quantifies the asymptotic deviation from the unquantized linear model. We also extend analytical techniques for stochastic gradient descent to nonlinear transformations of weights and inputs.

📄 PDF Abstract BibTeX arXiv:2510.10693

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

ProxQuant: Quantized Neural Networks via Proximal Operators

2018-10-01 · ICLR 2019 5 · Yu Bai, Yu-Xiang Wang, Edo Liberty

To make deep neural networks feasible in resource-constrained environments (such as mobile devices), it is beneficial to quantize models by using low-precision weights. One common technique for quantizing neural networks…

Quantization

Direct Quantized Training of Language Models with Stochastic Rounding

2024-12-06 · Kaiyan Zhao, Tsuguchika Tabaru, Kenichi Kobayashi, Takumi Honda 외

Although recent quantized Large Language Models (LLMs), such as BitNet, have paved the way for significant reduction in memory usage during deployment with binary or ternary weights, training these models still demands s…

iQRL -- Implicitly Quantized Representations for Sample-efficient Reinforcement Learning

2024-06-04 · Aidan Scannell, Kalle Kujanpää, Yi Zhao, Mohammadreza Nakhaei 외

Learning representations for reinforcement learning (RL) has shown much promise for continuous control. We propose an efficient representation learning method using only a self-supervised latent-state consistency loss. O…

continuous-controlContinuous Controlreinforcement-learningReinforcement Learning+2

Beyond Discreteness: Finite-Sample Analysis of Straight-Through Estimator for Quantization

2025-05-23 · Halyun Jeong, Jack Xin, Penghang Yin

Training quantized neural networks requires addressing the non-differentiable and discrete nature of the underlying optimization problem. To tackle this challenge, the straight-through estimator (STE) has become the most…

compressed sensingQuantization

Improving the Straight-Through Estimator with Zeroth-Order Information

2025-10-27 · Ningfeng Yang, Tor M. Aamodt arxiv

We study the problem of training neural networks with quantized parameters. Learning low-precision quantized parameters by enabling computation of gradients via the Straight-Through Estimator (STE) can be challenging. Wh…