Communication-Efficient Distributed SGD with Error-Feedback, Revisited
Ton Duc Thang University（孙德胜大学）；Meiji University（明治大学）；NICT（日本国立信息通信技术研究所）；National Institute of Information and Communications Technology (NICT)（日本国立信息通信技术研究所）
Tran Thi Phuong, Le Trieu Phong
发表信息：
International Journal of Computational Intelligence Systems, Volume 14, Issue 1, 2021, Pages 1373 -- 1387
We show that the convergence proof of a recent algorithm called dist-EF-SGD for distributed stochastic gradient descent with communication efficiency using error-feedback of Zheng et al. (NeurIPS 2019) is problematic mathematically. Concretely, the original error bound for arbitrary sequences of learning rate is unfortunately incorrect, leading to an invalidated upper bound in the convergence theorem for the algorithm. As evidences, we explicitly provide several counter-examples, for both convex and non-convex cases, to show the incorrectness of the error bound. We fix the issue by providing a new error bound and its corresponding proof, leading to a new convergence theorem for the dist-EF-SGD algorithm, and therefore recovering its mathematical analysis.
NTT DATA Mathematical Systems Inc.（NTT数据数学系统公司）；The University of Tokyo（东京大学）；RIKEN（理化学研究所）
A stochastic gradient method for synchronous distributed optimization is studied. For reducing communication cost, we particularly focus on utilization of compression of communicated gradients. Several work has shown that {\it{sparsified}} stochastic gradient descent method (SGD) with {\it{error feedback}} asymptotically achieves the same rate as (non-sparsified) parallel SGD. However, from a viewpoint of non-asymptotic behavior, the compression error may cause slower convergence than non-sparsified SGD in early iterations. This is problematic in practical situations since early stopping is of…
In this paper, we propose a unified analysis of variants of distributed SGD with arbitrary compressions and delayed updates. Our framework is general enough to cover different variants of quantized SGD, Error-Compensated SGD (EC-SGD) and SGD with delayed updates (D-SGD). Via a single theorem, we derive the complexity results for all the methods that fit our framework. For the existing methods, this theorem gives the best-known complexity results. Moreover, using our general scheme, we develop new variants of SGD that combine variance reduction or arbitrary sampling with error feedback and quan…
Sign-based algorithms (e.g. signSGD) have been proposed as a biased gradient compression technique to alleviate the communication bottleneck in training large neural networks across multiple workers. We show simple convex counter-examples where signSGD does not converge to the optimum. Further, even when it does converge, signSGD may generalize poorly when compared with SGD. These issues arise because of the biased nature of the sign compression operator. We then show that using error-feedback, i.e. incorporating the error made by the compression operator into the next step, overcomes these is…
The communication bottleneck has been a critical problem in large-scale distributed deep learning. In this work, we study distributed SGD with random block-wise sparsification as the gradient compressor, which is ring-allreduce compatible and highly computation-efficient but leads to inferior performance. To tackle this important issue, we improve the communication-efficient distributed SGD from a novel aspect, that is, the trade-off between the variance and second moment of the gradient. With this motivation, we propose a new detached error feedback (DEF) algorithm, which shows better converg…
Research Center for Applied Mathematics and Machine Intelligence（应用数学与机器智能研究中心）；Zhejiang Lab（之江实验室）
Averaging neural network weights sampled by a backbone stochastic gradient descent (SGD) is a simple yet effective approach to assist the backbone SGD in finding better optima, in terms of generalization. From a statistical perspective, weight averaging (WA) contributes to variance reduction. Recently, a well-established stochastic weight averaging (SWA) method is proposed, which is featured by the application of a cyclical or high constant (CHC) learning rate schedule (LRS) in generating weight samples for WA. Then a new insight on WA appears, which states that WA helps to discover wider opti…
We analyze (stochastic) gradient descent (SGD) with delayed updates on smooth quasi-convex and non-convex functions and derive concise, non-asymptotic, convergence rates. We show that the rate of convergence in all cases consists of two terms: (i) a stochastic term which is not affected by the delay, and (ii) a higher order deterministic term which is only linearly slowed down by the delay. Thus, in the presence of noise, the effects of the delay become negligible after a few iterations and the algorithm converges at the same optimal rate as standard SGD. This result extends a line of research…
Amazon Web Services（亚马逊云科技）；Hong Kong University of Science and Technology（香港科技大学）
Communication overhead is a major bottleneck hampering the scalability of distributed machine learning systems. Recently, there has been a surge of interest in using gradient compression to improve the communication efficiency of distributed neural network training. Using 1-bit quantization, signSGD with majority vote achieves a 32x reduction on communication cost. However, its convergence is based on unrealistic assumptions and can diverge in practice. In this paper, we propose a general distributed compressed SGD with Nesterov's momentum. We consider two-way compression, which compresses the…
Department of Computer Science and Engineering（计算机科学与工程系）；State University of New York at Buffalo（纽约州立大学布法罗分校）
In this paper, we revisit the large-scale constrained linear regression problem and propose faster methods based on some recent developments in sketching and optimization. Our algorithms combine (accelerated) mini-batch SGD with a new method called two-step preconditioning to achieve an approximate solution with a time complexity lower than that of the state-of-the-art techniques for the low precision case. Our idea can also be extended to the high precision case, which gives an alternative implementation to the Iterative Hessian Sketch (IHS) method with significantly improved time complexity.…
Stochastic Gradient Descent (SGD) is arguably the most popular of the machine learning methods applied to training deep neural networks (DNN) today. It has recently been demonstrated that SGD can be statistically biased so that certain elements of the training set are learned more rapidly than others. In this article, we place SGD into a feedback loop whereby the probability of selection is proportional to error magnitude. This provides a novelty-driven oddball SGD process that learns more rapidly than traditional SGD by prioritising those elements of the training set with the largest novelty …
ETH Zürich（苏黎世联邦理工学院）；University of Cambridge（剑桥大学）；Google AI Berlin（谷歌柏林人工智能研究院）；Imperial College London（伦敦帝国理工学院）；University of Bristol（布里斯托尔大学）
Isotropic Gaussian priors are the de facto standard for modern Bayesian neural network inference. However, it is unclear whether these priors accurately reflect our true beliefs about the weight distributions or give optimal performance. To find better priors, we study summary statistics of neural network weights in networks trained using stochastic gradient descent (SGD). We find that convolutional neural network (CNN) and ResNet weights display strong spatial correlations, while fully connected networks (FCNNs) display heavy-tailed weight distributions. We show that building these observatio…
An increasing bottleneck in decentralized optimization is communication. Bigger models and growing datasets mean that decentralization of computation is important and that the amount of information exchanged is quickly growing. While compression techniques have been introduced to cope with the latter, none has considered leveraging the temporal correlations that exist in consecutive vector updates. An important example is distributed momentum-SGD where temporal correlation is enhanced by the low-pass-filtering effect of applying momentum. In this paper we design and analyze compression methods…
IST Austria（奥地利科学技术研究所）；Czech Technical University in Prague（布拉格捷克理工大学）
Machine learning has made tremendous progress in recent years, with models matching or even surpassing humans on a series of specialized tasks. One key element behind the progress of machine learning in recent years has been the ability to train machine learning models in large-scale distributed shared-memory and message-passing environments. Many of these models are trained employing variants of stochastic gradient descent (SGD) based optimization. In this paper, we introduce a general consistency condition covering communication-reduced and asynchronous distributed SGD implementations. Our f…
Lossy gradient compression, with either unbiased or biased compressors, has become a key tool to avoid the communication bottleneck in centrally coordinated distributed training of machine learning models. We analyze the performance of two standard and general types of methods: (i) distributed quantized SGD (D-QSGD) with arbitrary unbiased quantizers and (ii) distributed SGD with error-feedback and biased compressors (D-EF-SGD) in the heterogeneous (non-iid) data setting. Our results indicate that D-EF-SGD is much less affected than D-QSGD by non-iid data, but both methods can suffer a slowdow…
Accelerated zero-order SGD under high-order smoothness and overparameterized regime
Lomonosov MSU（罗蒙诺索夫莫斯科大学）；ISP RAS（俄罗斯科学院信息问题研究所）；HSE University（高等经济大学）；MIPT（莫斯科物理技术学院）；MSU Institute for Artificial Intelligence（莫斯科大学人工智能研究所）；Innopolis University（因诺波利斯大学）；IITP RAS（俄罗斯科学院信息传输问题研究所）；Skoltech（斯科尔科沃科学技术研究院）
Center for Brains, Minds and Machines, McGovern Institute（麦戈文研究所脑、心智与机器中心）；Massachusetts Institute of Technology（麻省理工学院）
Gradient backpropagation (BP) requires symmetric feedforward and feedback connections -- the same weights must be used for forward and backward passes. This "weight transport problem" (Grossberg 1987) is thought to be one of the main reasons to doubt BP's biologically plausibility. Using 15 different classification datasets, we systematically investigate to what extent BP really depends on weight symmetry. In a study that turned out to be surprisingly similar in spirit to Lillicrap et al.'s demonstration (Lillicrap et al. 2014) but orthogonal in its results, our experiments indicate that: (1) …
Bytedance AI Lab（字节跳动人工智能实验室）；Peking University（北京大学）；Ohio State University（俄亥俄州立大学）；Beijing Institute of Technology（北京理工大学）
Due to its simplicity and outstanding ability to generalize, stochastic gradient descent (SGD) is still the most widely used optimization method despite its slow convergence. Meanwhile, adaptive methods have attracted rising attention of optimization and machine learning communities, both for the leverage of life-long information and for the profound and fundamental mathematical theory. Taking the best of both worlds is the most exciting and challenging question in the field of optimization for machine learning. Along this line, we revisited existing adaptive gradient methods from a novel pers…
University of Konstanz（康斯坦茨大学）；National University of Singapore（新加坡国立大学）；Google Inc.（谷歌公司）
In recent years, interest in recommender research has shifted from explicit feedback towards implicit feedback data. A diversity of complex models has been proposed for a wide variety of applications. Despite this, learning from implicit feedback is still computationally challenging. So far, most work relies on stochastic gradient descent (SGD) solvers which are easy to derive, but in practice challenging to apply, especially for tasks with many items. For the simple matrix factorization model, an efficient coordinate descent (CD) solver has been previously proposed. However, efficient CD appr…
The synthesis problem of static output feedback controllers within the anistropic-norm setup is revisited. A tractable synthesis approach involving iterations over a convex optimization problem is suggested, similarly to existing results for the Hinfinity-norm minimization case. The results are formulated by a couple of Linear Matrix Inequalities coupled via a bilinear equality, revealing, as in the Hinfinity case, the duality between the control-type and filtering type LMIs and allowing a tractable iterative method to cope with practical static output feedback synthesis problems. The resultin…
University of Amsterdam（阿姆斯特丹大学）；Ahold Delhaize（阿霍德德尔海兹）
Counterfactual Learning to Rank (LTR) algorithms learn a ranking model from logged user interactions, often collected using a production system. Employing such an offline learning approach has many benefits compared to an online one, but it is challenging as user feedback often contains high levels of bias. Unbiased LTR uses Inverse Propensity Scoring (IPS) to enable unbiased learning from logged user interactions. One of the major difficulties in applying Stochastic Gradient Descent (SGD) approaches to counterfactual learning problems is the large variance introduced by the propensity weights…
In federated learning (FL) systems, e.g., wireless networks, the communication cost between the clients and the central server can often be a bottleneck. To reduce the communication cost, the paradigm of communication compression has become a popular strategy in the literature. In this paper, we focus on biased gradient compression techniques in non-convex FL problems. In the classical setting of distributed learning, the method of error feedback (EF) is a common technique to remedy the downsides of biased gradient compression. In this work, we study a compressed FL scheme equipped with error …
Double Momentum and Error Feedback for Clipping with Fast Rates and Differential Privacy
University of Basel（巴塞尔大学）；Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)（穆罕默德·本·扎耶德人工智能大学（MBZUAI））；King Abdullah University of Science and Technology (KAUST)（阿卜杜拉国王科技大学（KAUST））
Reducing the variance of the gradient estimator is known to improve the convergence rate of stochastic gradient-based optimization and sampling algorithms. One way of achieving variance reduction is to design importance sampling strategies. Recently, the problem of designing such schemes was formulated as an online learning problem with bandit feedback, and algorithms with sub-linear static regret were designed. In this work, we build on this framework and propose Avare, a simple and efficient algorithm for adaptive importance sampling for finite-sum optimization and sampling with decreasing s…
Gradient compression is a widely-established remedy to tackle the communication bottleneck in distributed training of large deep neural networks (DNNs). Under the error-feedback framework, Top-$k$ sparsification, sometimes with $k$ as little as $0.1\%$ of the gradient size, enables training to the same model quality as the uncompressed case for a similar iteration count. From the optimization perspective, we find that Top-$k$ is the communication-optimal sparsifier given a per-iteration $k$ element budget. We argue that to further the benefits of gradient sparsification, especially for DNNs, a…
We develop a model-free learning algorithm for the infinite-horizon linear quadratic regulator (LQR) problem. Specifically, (risk) constraints and structured feedback are considered, in order to reduce the state deviation while allowing for a sparse communication graph in practice. By reformulating the dual problem as a nonconvex-concave minimax problem, we adopt the gradient descent max-oracle (GDmax), and for modelfree setting, the stochastic (S)GDmax using zero-order policy gradient. By bounding the Lipschitz and smoothness constants of the LQR cost using specifically defined sublevel sets,…
This paper investigates stochastic and adversarial combinatorial multi-armed bandit problems. In the stochastic setting under semi-bandit feedback, we derive a problem-specific regret lower bound, and discuss its scaling with the dimension of the decision space. We propose ESCB, an algorithm that efficiently exploits the structure of the problem and provide a finite-time analysis of its regret. ESCB has better performance guarantees than existing algorithms, and significantly outperforms these algorithms in practice. In the adversarial setting under bandit feedback, we propose \textsc{CombEXP}…
Pennsylvania State University（宾夕法尼亚州立大学）；University of Virginia（弗吉尼亚大学）
Due to the explosion in the size of the training datasets, distributed learning has received growing interest in recent years. One of the major bottlenecks is the large communication cost between the central server and the local workers. While error feedback compression has been proven to be successful in reducing communication costs with stochastic gradient descent (SGD), there are much fewer attempts in building communication-efficient adaptive gradient methods with provable guarantees, which are widely used in training large-scale machine learning models. In this paper, we propose a new com…
In this paper, we revisit the problem of finding the average capacity of the Gaussian feedback channel. First, we consider the problem of finding the average capacity of the analog Gaussian noise channel where the noise has an arbitrary spectral density. We introduce a new approach to the problem where we solve the problem over a finite number of transmissions and then consider the limit of an infinite number of transmissions. Further, we consider the important special case of stationary Gaussian noise with finite memory. We show that the channel capacity at stationarity can be found by solvin…
Gradient compression is a recent and increasingly popular technique for reducing the communication cost in distributed training of large-scale machine learning models. In this work we focus on developing efficient distributed methods that can work for any compressor satisfying a certain contraction property, which includes both unbiased (after appropriate scaling) and biased compressors such as RandK and TopK. Applied naively, gradient compression introduces errors that either slow down convergence or lead to divergence. A popular technique designed to tackle this issue is error compensation/e…
King Abdullah University of Science and Technology（阿卜杜拉国王科技大学）；Skolkovo Institute of Science and Technology（斯科尔科沃科学技术研究院）；Moscow Institute of Physics and Technology（莫斯科物理技术学院）；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）；Institute of Science and Technology Austria（奥地利科学技术研究院）
Load frequency control (LFC) is a key factor to maintain the stable frequency in multi-area power systems. As the modern power systems evolve from centralized to distributed paradigm, LFC needs to consider the peer-to-peer (P2P) based scheme that considers limited information from the information-exchange graph for the generator control of each interconnected area. This paper aims to solve a data-driven constrained LQR problem with mean-variance risk constraints and output structured feedback, and applies this framework to solve the LFC problem in multi-area power systems. By reformulating the…
SA-PEF: Step-Ahead Partial Error Feedback for Efficient Federated Learning
Department of Electronic Systems, Norwegian University of Science and Technology (NTNU)（挪威科学技术大学电子系统系）；Department of Information and Communications Engineering, Aalto University（阿尔托大学信息与通信工程系）；CSIRO’s Data61（CSIRO数据61）
