Custom Gradient Estimators are Straight-Through Estimators in Disguise

• URL: http://arxiv.org/abs/2405.05171v3
• Date: Wed, 22 May 2024 19:36:06 GMT
• Title: Custom Gradient Estimators are Straight-Through Estimators in Disguise
• Authors: Matt Schoenbauer, Daniele Moro, Lukasz Lew, Andrew Howard,
• Abstract summary: Quantization-aware training comes with a fundamental challenge: the derivative of quantization functions such as rounding are zero almost everywhere. We prove that when the learning rate is sufficiently small, a large class of weight gradient estimators is equivalent with the straight through estimator (STE) We experimentally show that these results hold for both a small convolutional model trained on the MNIST dataset and for a ResNet50 model trained on ImageNet.
• Score: 3.1037083241174197
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Quantization-aware training comes with a fundamental challenge: the derivative of quantization functions such as rounding are zero almost everywhere and nonexistent elsewhere. Various differentiable approximations of quantization functions have been proposed to address this issue. In this paper, we prove that when the learning rate is sufficiently small, a large class of weight gradient estimators is equivalent with the straight through estimator (STE). Specifically, after swapping in the STE and adjusting both the weight initialization and the learning rate in SGD, the model will train in almost exactly the same way as it did with the original gradient estimator. Moreover, we show that for adaptive learning rate algorithms like Adam, the same result can be seen without any modifications to the weight initialization and learning rate. We experimentally show that these results hold for both a small convolutional model trained on the MNIST dataset and for a ResNet50 model trained on ImageNet.

Related papers

• Training trajectories, mini-batch losses and the curious role of the learning rate [13.848916053916618]
  We show that validated gradient descent plays a fundamental role in nearly all applications of deep learning. We propose a simple model and a geometric interpretation that allows to analyze the relationship between the gradients of mini-batches and the full batch. In particular, a very low loss value can be reached just one step of descent with large enough learning rate.
  arXiv  Detail & Related papers  (2023-01-05T21:58:46Z)
• 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)
• Training Discrete Deep Generative Models via Gapped Straight-Through Estimator [72.71398034617607]
  We propose a Gapped Straight-Through ( GST) estimator to reduce the variance without incurring resampling overhead. This estimator is inspired by the essential properties of Straight-Through Gumbel-Softmax. Experiments demonstrate that the proposed GST estimator enjoys better performance compared to strong baselines on two discrete deep generative modeling tasks.
  arXiv  Detail & Related papers  (2022-06-15T01:46:05Z)
• Merging Models with Fisher-Weighted Averaging [24.698591753644077]
  We introduce a fundamentally different method for transferring knowledge across models that amounts to "merging" multiple models into one. Our approach effectively involves computing a weighted average of the models' parameters. We show that our merging procedure makes it possible to combine models in previously unexplored ways.
  arXiv  Detail & Related papers  (2021-11-18T17:59:35Z)
• Unrolling SGD: Understanding Factors Influencing Machine Unlearning [17.6607904333012]
  Machine unlearning is the process through which a deployed machine learning model forgets about one of its training data points. We first taxonomize approaches and metrics of approximate unlearning. We identify verification error, i.e., the L2 difference between the weights of an approximately unlearned and a naively retrained model.
  arXiv  Detail & Related papers  (2021-09-27T23:46:59Z)
• 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)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)
• Path Sample-Analytic Gradient Estimators for Stochastic Binary Networks [78.76880041670904]
  In neural networks with binary activations and or binary weights the training by gradient descent is complicated. We propose a new method for this estimation problem combining sampling and analytic approximation steps. We experimentally show higher accuracy in gradient estimation and demonstrate a more stable and better performing training in deep convolutional models.
  arXiv  Detail & Related papers  (2020-06-04T21:51:21Z)
• Dynamic Scale Training for Object Detection [111.33112051962514]
  We propose a Dynamic Scale Training paradigm (abbreviated as DST) to mitigate scale variation challenge in object detection. Experimental results demonstrate the efficacy of our proposed DST towards scale variation handling. It does not introduce inference overhead and could serve as a free lunch for general detection configurations.
  arXiv  Detail & Related papers  (2020-04-26T16:48:17Z)

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.