Improving the Straight-Through Estimator with Zeroth-Order Information

• URL: http://arxiv.org/abs/2510.23926v1
• Date: Mon, 27 Oct 2025 23:14:59 GMT
• Title: Improving the Straight-Through Estimator with Zeroth-Order Information
• Authors: Ningfeng Yang, Tor M. Aamodt,
• Abstract summary: We study the problem of training neural networks with quantized parameters.<n>We propose First-Order-Guided Zeroth-Order Gradient Descent (FOGZO)<n>We show FOGZO improves the tradeoff between quality and training time in Quantization-Aware Pre-Training.
• Score: 7.09016563801433
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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. While the STE enables back-propagation, which is a first-order method, recent works have explored the use of zeroth-order (ZO) gradient descent for fine-tuning. We note that the STE provides high-quality biased gradients, and ZO gradients are unbiased but can be expensive. We thus propose First-Order-Guided Zeroth-Order Gradient Descent (FOGZO) that reduces STE bias while reducing computations relative to ZO methods. Empirically, we show FOGZO improves the tradeoff between quality and training time in Quantization-Aware Pre-Training. Specifically, versus STE at the same number of iterations, we show a 1-8\% accuracy improvement for DeiT Tiny/Small, 1-2\% accuracy improvement on ResNet 18/50, and 1-22 perplexity point improvement for LLaMA models with up to 0.3 billion parameters. For the same loss, FOGZO yields a 796$\times$ reduction in computation versus n-SPSA for a 2-layer MLP on MNIST. Code is available at https://github.com/1733116199/fogzo.

Related papers

• QuZO: Quantized Zeroth-Order Fine-Tuning for Large Language Models [27.730213115659986]
  Language Models (LLMs) are often quantized to lower precision to reduce the memory cost and latency in inference.<n>Traditional fine-tuning methods require backpropagation, which are error-prone in the low-precision settings.<n>We propose the Quantized Zeroth-Order (QuZO) framework, specifically designed for fine-tuning LLMs through low-precision forward passes.
  arXiv  Detail & Related papers  (2025-02-17T22:20:31Z)
• Zeroth-Order Fine-Tuning of LLMs in Random Subspaces [63.10833446782114]
  As language models grow in size, memory demands for backpropagation increase.<n>Zeroth-order (ZO) optimization methods offer a memory-efficient alternative.<n>In this paper, we propose Subspace Zero-order optimization to address the challenges posed by posed by high dimensionality perturbations.
  arXiv  Detail & Related papers  (2024-10-11T17:01:43Z)
• Sparse is Enough in Fine-tuning Pre-trained Large Language Models [98.46493578509039]
  We propose a gradient-based sparse fine-tuning algorithm, named Sparse Increment Fine-Tuning (SIFT) We validate its effectiveness on a range of tasks including the GLUE Benchmark and Instruction-tuning.
  arXiv  Detail & Related papers  (2023-12-19T06:06:30Z)
• Relationship between Batch Size and Number of Steps Needed for Nonconvex Optimization of Stochastic Gradient Descent using Armijo Line Search [0.8158530638728501]
  We show that SGD performs better than other deep learning networks when it uses deep numerical line. The results indicate that the number of steps needed for SFO as the batch size grows can be estimated.
  arXiv  Detail & Related papers  (2023-07-25T21:59:17Z)
• Fighting Uncertainty with Gradients: Offline Reinforcement Learning via Diffusion Score Matching [22.461036967440723]
  We study smoothed distance to data as an uncertainty metric, and claim that it has two beneficial properties. We show these gradients can be efficiently learned with score-matching techniques. We propose Score-Guided Planning (SGP) to enable first-order planning in high-dimensional problems.
  arXiv  Detail & Related papers  (2023-06-24T23:40:58Z)
• Scaling Forward Gradient With Local Losses [117.22685584919756]
  Forward learning is a biologically plausible alternative to backprop for learning deep neural networks. We show that it is possible to substantially reduce the variance of the forward gradient by applying perturbations to activations rather than weights. Our approach matches backprop on MNIST and CIFAR-10 and significantly outperforms previously proposed backprop-free algorithms on ImageNet.
  arXiv  Detail & Related papers  (2022-10-07T03:52:27Z)
• Tom: Leveraging trend of the observed gradients for faster convergence [0.0]
  Tom is a novel variant of Adam that takes into account the trend observed for the gradients in the landscape in the loss traversed by the neural network. Tom outperforms Adagrad, Adadelta, RMSProp and Adam in terms of both accuracy and has a faster convergence.
  arXiv  Detail & Related papers  (2021-09-07T20:19:40Z)
• Differentiable Annealed Importance Sampling and the Perils of Gradient Noise [68.44523807580438]
  Annealed importance sampling (AIS) and related algorithms are highly effective tools for marginal likelihood estimation. Differentiability is a desirable property as it would admit the possibility of optimizing marginal likelihood as an objective. We propose a differentiable algorithm by abandoning Metropolis-Hastings steps, which further unlocks mini-batch computation.
  arXiv  Detail & Related papers  (2021-07-21T17:10:14Z)
• Exploiting Adam-like Optimization Algorithms to Improve the Performance of Convolutional Neural Networks [82.61182037130405]
  gradient descent (SGD) is the main approach for training deep networks. In this work, we compare Adam based variants based on the difference between the present and the past gradients. We have tested ensemble of networks and the fusion with ResNet50 trained with gradient descent.
  arXiv  Detail & Related papers  (2021-03-26T18:55:08Z)
• Correcting Momentum with Second-order Information [50.992629498861724]
  We develop a new algorithm for non-critical optimization that finds an $O(epsilon)$epsilon point in the optimal product. We validate our results on a variety of large-scale deep learning benchmarks and architectures.
  arXiv  Detail & Related papers  (2021-03-04T19:01:20Z)
• Carath\'eodory Sampling for Stochastic Gradient Descent [79.55586575988292]
  We present an approach that is inspired by classical results of Tchakaloff and Carath'eodory about measure reduction. We adaptively select the descent steps where the measure reduction is carried out. We combine this with Block Coordinate Descent so that measure reduction can be done very cheaply.
  arXiv  Detail & Related papers  (2020-06-02T17:52:59Z)
• Resolving learning rates adaptively by locating Stochastic Non-Negative Associated Gradient Projection Points using line searches [0.0]
  Learning rates in neural network training are currently determined a priori to training using expensive manual or automated tuning. This study proposes gradient-only line searches to resolve the learning rate for neural network training algorithms.
  arXiv  Detail & Related papers  (2020-01-15T03:08:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.