A Scalable Finite Difference Method for Deep Reinforcement Learning

• URL: http://arxiv.org/abs/2210.07487v1
• Date: Fri, 14 Oct 2022 03:33:53 GMT
• Title: A Scalable Finite Difference Method for Deep Reinforcement Learning
• Authors: Matthew Allen, John Raisbeck, and Hakho Lee
• Abstract summary: We investigate a problem with the use of distributed workers in some Deep Reinforcement Learning domains. We produce a stable, low-bandwidth learning algorithm that achieves 100% usage of all connected CPUs under typical conditions.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Several low-bandwidth distributable black-box optimization algorithms have recently been shown to perform nearly as well as more refined modern methods in some Deep Reinforcement Learning domains. In this work we investigate a core problem with the use of distributed workers in such systems. Further, we investigate the dramatic differences in performance between the popular Adam gradient descent algorithm and the simplest form of stochastic gradient descent. These investigations produce a stable, low-bandwidth learning algorithm that achieves 100\% usage of all connected CPUs under typical conditions.

Related papers

• A Mirror Descent-Based Algorithm for Corruption-Tolerant Distributed Gradient Descent [57.64826450787237]
  We show how to analyze the behavior of distributed gradient descent algorithms in the presence of adversarial corruptions. We show how to use ideas from (lazy) mirror descent to design a corruption-tolerant distributed optimization algorithm. Experiments based on linear regression, support vector classification, and softmax classification on the MNIST dataset corroborate our theoretical findings.
  arXiv  Detail & Related papers  (2024-07-19T08:29:12Z)
• GRAWA: Gradient-based Weighted Averaging for Distributed Training of Deep Learning Models [9.377424534371727]
  We study distributed training of deep models in time-constrained environments. We propose a new algorithm that periodically pulls workers towards the center variable computed as an average of workers.
  arXiv  Detail & Related papers  (2024-03-07T04:22:34Z)
• Ordering for Non-Replacement SGD [7.11967773739707]
  We seek to find an ordering that can improve the convergence rates for the non-replacement form of the algorithm. We develop optimal orderings for constant and decreasing step sizes for strongly convex and convex functions. In addition, we are able to combine the ordering with mini-batch and further apply it to more complex neural networks.
  arXiv  Detail & Related papers  (2023-06-28T00:46:58Z)
• A Bootstrap Algorithm for Fast Supervised Learning [0.0]
  Training a neural network (NN) typically relies on some type of curve-following method, such as gradient descent (and gradient descent (SGD)), ADADELTA, ADAM or limited memory algorithms. Convergence for these algorithms usually relies on having access to a large quantity of observations in order to achieve a high level of accuracy and, with certain classes of functions, these algorithms could take multiple epochs of data points to catch on. Herein, a different technique with the potential of achieving dramatically better speeds of convergence is explored: it does not curve-follow but rather relies on 'decoupling' hidden layers and on
  arXiv  Detail & Related papers  (2023-05-04T18:28:18Z)
• The Cascaded Forward Algorithm for Neural Network Training [61.06444586991505]
  We propose a new learning framework for neural networks, namely Cascaded Forward (CaFo) algorithm, which does not rely on BP optimization as that in FF. Unlike FF, our framework directly outputs label distributions at each cascaded block, which does not require generation of additional negative samples. In our framework each block can be trained independently, so it can be easily deployed into parallel acceleration systems.
  arXiv  Detail & Related papers  (2023-03-17T02:01:11Z)
• 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)
• Simple Stochastic and Online Gradient DescentAlgorithms for Pairwise Learning [65.54757265434465]
  Pairwise learning refers to learning tasks where the loss function depends on a pair instances. Online descent (OGD) is a popular approach to handle streaming data in pairwise learning. In this paper, we propose simple and online descent to methods for pairwise learning.
  arXiv  Detail & Related papers  (2021-11-23T18:10:48Z)
• Evolving Reinforcement Learning Algorithms [186.62294652057062]
  We propose a method for meta-learning reinforcement learning algorithms. The learned algorithms are domain-agnostic and can generalize to new environments not seen during training. We highlight two learned algorithms which obtain good generalization performance over other classical control tasks, gridworld type tasks, and Atari games.
  arXiv  Detail & Related papers  (2021-01-08T18:55:07Z)
• Multi-kernel Passive Stochastic Gradient Algorithms and Transfer Learning [21.796874356469644]
  The gradient algorithm does not have control over the location where noisy gradients of the cost function are evaluated. The algorithm performs substantially better in high dimensional problems and incorporates variance reduction.
  arXiv  Detail & Related papers  (2020-08-23T11:55:19Z)
• Differentially Private Accelerated Optimization Algorithms [0.7874708385247353]
  We present two classes of differentially private optimization algorithms. The first algorithm is inspired by Polyak's heavy ball method. The second class of algorithms are based on Nesterov's accelerated gradient method.
  arXiv  Detail & Related papers  (2020-08-05T08:23:01Z)
• Variance Reduction with Sparse Gradients [82.41780420431205]
  Variance reduction methods such as SVRG and SpiderBoost use a mixture of large and small batch gradients. We introduce a new sparsity operator: The random-top-k operator. Our algorithm consistently outperforms SpiderBoost on various tasks including image classification, natural language processing, and sparse matrix factorization.
  arXiv  Detail & Related papers  (2020-01-27T08:23:58Z)

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.