Distributed Machine Learning (DML) is proposed not only to accelerate the training of machine learning, but also to solve the inadequate ability for handling a large amount of training data. It adopts multiple computing nodes in data center to collaboratively work in parallel at the cost of high communication overhead. Gradient Compression (GC) is introduced to reduce the communication overhead by reducing the number of synchronized gradients among computing nodes. However, existing GC solutions suffer from varying network congestion. To be specific, when some computing nodes experience high network congestion, their gradient transmission process could be significantly delayed, slowing down the entire training process. To solve the problem, we propose FLASH, a congestion-aware GC solution for DML. FLASH accelerates the training process by jointly considering the iterative approximation of machine learning and dynamic network congestion scenarios. It can maintain good training performance by adaptively adjust and schedule the number of synchronized gradients among computing nodes. We evaluate the effectiveness of FLASH using AlexNet and Resnet18 under different network congestion scenarios. Simulation results show that under the same number of training epochs, FLASH reduces training time 22-71%, maintains good accuracy, and low loss, compared with the existing memory top-K GC solution.
2 March 2020 in IEEE Transactions on Signal Processing
This paper proposes a gradient-based multi-agent actor-critic algorithm for off-policy reinforcement learning using importance sampling. Our algorithm is incremental with full gradients, and its complexity per iteration scales linearly with the size of approximation features. Previous multi-agent actor-critic algorithms are limited to the on-policy setting or off-policy emphatic temporal difference (TD) learning and they do not take advantage of the advances in off-policy gradient temporal difference learning (GTD). As a theoretical contribution, we establish that the critic step of the proposed algorithm converges to the TD solution of the projected Bellman equation and the actor step converges to the set of asymptotically stable fixed points. Numerical experiments on the multi-agent generalization of the Boyan’s chain problem show that the proposed approach provides improved performances in terms of stability and convergence rate as compared with the state-of-the-art baseline algorithm.
18 December 2014 in Journal of Inequalities and Applications
In this paper, we study the properties of a bilinear multiplier space. We give a necessary condition for a continuous bounded function to be a bilinear multiplier on variable exponent Lebesgue spaces, and we prove the localization theorem of multipliers on variable exponent Lebesgue spaces. Moreover, we present a Mihlin-Hörmander type theorem for multilinear Fourier multipliers on weighted variable Lebesgue spaces and give some applications.
21 May 2020 in IEEE Transactions on Signal Processing
This paper proposes and analyzes a communication-efficient distributed optimization framework for general nonconvex nonsmooth signal processing and machine learning problems under an asynchronous protocol. At each iteration, worker machines compute gradients of a known empirical loss function using their own local data, and a master machine solves a related minimization problem to update the current estimate. We prove that for nonconvex nonsmooth problems, the proposed algorithm converges to a stationary point with a sublinear rate over the number of communication rounds, coinciding with the best theoretical rate that can be achieved for this class of problems. Linear convergence to a global minimum is established without any statistical assumptions on the local data for problems characterized by composite loss functions whose smooth part is strongly convex. Extensive numerical experiments verify that the performance of the proposed approach indeed improves -- sometimes significantly -- over other state-of-the-art algorithms in terms of total communication efficiency.
25 February 2022 in IEEE Transactions on Mobile Computing
Edge-assisted Distributed Learning (EDL) is a popular machine learning paradigm that uses a set of distributed edge nodes to collaboratively train a machine learning model using training data. Most of existing works implicitly assume that the fixed amount of training data is pre-collected and dispatched from user devices to edge nodes. In real world, however, training data in edge nodes are collected from user devices through wireless networks, and the volume and distribution of training data in edge nodes could exhibit temporal and spatial fluctuations due to varying wireless situations (e.g., network congestion, link capacity variation). In this way, existing solutions suffer from slow convergence and low accuracy. In this paper, we propose SpongeTraining to achieve high efficiency and accuracy for online EDL. To accommodate to fluctuations in training data, SpongeTraining uses a buffer at each worker to store received training data and adaptively adjusts training batch size and learning rate of each worker based on training data extracted from the buffer. Experiment results based on real-world datasets show that SpongeTraining outperforms existing solutions by accelerating the training process up to 50% for reaching the same training accuracy.