skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Taming the Noisy Gradient: Train Deep Neural Networks with Small Batch Sizes
Deep learning architectures are usually proposed with millions of parameters, resulting in a memory issue when training deep neural networks with stochastic gradient descent type methods using large batch sizes. However, training with small batch sizes tends to produce low quality solution due to the large variance of stochastic gradients. In this paper, we tackle this problem by proposing a new framework for training deep neural network with small batches/noisy gradient. During optimization, our method iteratively applies a proximal type regularizer to make loss function strongly convex. Such regularizer stablizes the gradient, leading to better training performance. We prove that our algorithm achieves comparable convergence rate as vanilla SGD even with small batch size. Our framework is simple to implement and can be potentially combined with many existing optimization algorithms. Empirical results show that our method outperforms SGD and Adam when batch size is small. Our implementation is available at https://github.com/huiqu18/TRAlgorithm.  more » « less
Award ID(s):
1855759
PAR ID:
10108626
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
The Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI)
Page Range / eLocation ID:
4348 to 4354
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; (, Center for Brains, Minds and Machines (CBMM))
    We analyze deep ReLU neural networks trained with mini-batch stochastic gradient decent and weight decay. We prove that the source of the SGD noise is an implicit low rank constraint across all of the weight matrices within the network. Furthermore, we show, both theoretically and empirically, that when training a neural network using Stochastic Gradient Descent (SGD) with a small batch size, the resulting weight matrices are expected to be of small rank. Our analysis relies on a minimal set of assumptions and the neural networks may include convolutional layers, residual connections, as well as batch normalization layers. 
    more » « less
  2. ; ; ; (, Center for Brains, Minds and Machines (CBMM))
    In this paper, we study the bias of Stochastic Gradient Descent (SGD) to learn low-rank weight matrices when training deep ReLU neural networks. Our results show that training neural networks with mini-batch SGD and weight decay causes a bias towards rank minimization over the weight matrices. Specifically, we show, both theoretically and empirically, that this bias is more pronounced when using smaller batch sizes, higher learning rates, or increased weight decay. Additionally, we predict and observe empirically that weight decay is necessary to achieve this bias. Finally, we empirically investigate the connection between this bias and generalization, finding that it has a marginal effect on generalization. Our analysis is based on a minimal set of assumptions and applies to neural networks of any width or depth, including those with residual connections and convolutional layers. 
    more » « less
  3. ; (, IEEE International Conference on Cloud Engineering (IC2E))
    Machine learning models such as deep neural networks have been shown to be successful in solving a wide range of problems. Training such a model typically requires stochastic gradient descent, and the process is time-consuming and expensive in terms of computing resources. In this paper, we propose a distributed framework that supports the prioritized execution of the gradient computation. Our proposed distributed framework identifies important data points through computing or estimating the priority for each data point. We evaluate the proposed distributed framework with several machine learning models including multi-layer perceptron (MLP) and convolutional neural networks (CNN). Our experimental results show that prioritized SGD accelerates the training of machine learning models by as much as 1.6X over that of the mini-batch SGD. Further, the distributed framework scales linearly with the number of workers. 
    more » « less
  4. ; ; ; ; (, Journal of machine learning research)
    This work characterizes the benefits of averaging techniques widely used in conjunction with stochastic gradient descent (SGD). In particular, this work presents a sharp analysis of: (1) mini- batching, a method of averaging many samples of a stochastic gradient to both reduce the variance of a stochastic gradient estimate and for parallelizing SGD and (2) tail-averaging, a method involving averaging the final few iterates of SGD in order to decrease the variance in SGD’s final iterate. This work presents sharp finite sample generalization error bounds for these schemes for the stochastic approximation problem of least squares regression. Furthermore, this work establishes a precise problem-dependent extent to which mini-batching can be used to yield provable near-linear parallelization speedups over SGD with batch size one. This characterization is used to understand the relationship between learning rate versus batch size when considering the excess risk of the final iterate of an SGD procedure. Next, this mini-batching characterization is utilized in providing a highly parallelizable SGD method that achieves the min- imax risk with nearly the same number of serial updates as batch gradient descent, improving significantly over existing SGD-style methods. Following this, a non-asymptotic excess risk bound for model averaging (which is a communication efficient parallelization scheme) is provided. Finally, this work sheds light on fundamental differences in SGD’s behavior when dealing with mis-specified models in the non-realizable least squares problem. This paper shows that maximal stepsizes ensuring minimax risk for the mis-specified case must depend on the noise properties. The analysis tools used by this paper generalize the operator view of averaged SGD (De ́fossez and Bach, 2015) followed by developing a novel analysis in bounding these operators to char- acterize the generalization error. These techniques are of broader interest in analyzing various computational aspects of stochastic approximation. 
    more » « less
  5. ; (, IEEE International Symposium on Information Theory (ISIT))
    We study stochastic gradient descent (SGD) in the master-worker architecture under Byzantine attacks. Building upon the recent advances in algorithmic high-dimensional robust statistics, in each SGD iteration, master employs a non-trivial decoding to estimate the true gradient from the unbiased stochastic gradients received from workers, some of which may be corrupt. We provide convergence analyses for both strongly-convex and non-convex smooth objectives under standard SGD assumptions. We can control the approximation error of our solution in both these settings by the mini-batch size of stochastic gradients; and we can make the approximation error as small as we want, provided that workers use a sufficiently large mini-batch size. Our algorithm can tolerate less than 1/3 fraction of Byzantine workers. It can approximately find the optimal parameters in the strongly-convex setting exponentially fast, and reaches to an approximate stationary point in the non-convex setting with linear speed, i.e., with a rate of 1/T, thus, matching the convergence rates of vanilla SGD in the Byzantine-free setting. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email