arXiv subjects

Zhize Li

Publications and source records attributed to Zhize Li.

34 records · Page 2Linked to original sources

arXiv2021

FedPAGE: A Fast Local Stochastic Gradient Method for Communication-Efficient Federated Learning

Federated Averaging (FedAvg, also known as Local-SGD) (McMahan et al., 2017) is a classical federated learning algorithm in which clients run multiple local SGD steps before communicating their update to an orchestrating server. We propose a new federated learning algorithm, FedPAGE, able to further reduce the communication complexity by utilizing the recent optimal PAGE method (Li et al., 2021) instead of plain SGD in FedAvg. We show that FedPAGE uses much fewer communication rounds than previous local methods for both federated convex and nonconvex optimization. Concretely, 1) in the convex setting, the number of communication rounds of FedPAGE is $O(\frac{N^{3/4}}{Sε})$, improving the best-known result $O(\frac{N}{Sε})$ of SCAFFOLD (Karimireddy et al.,2020) by a factor of $N^{1/4}$, where $N$ is the total number of clients (usually is very large in federated learning), $S$ is the sampled subset of clients in each communication round, and $ε$ is the target error; 2) in the nonconvex setting, the number of communication rounds of FedPAGE is $O(\frac{\sqrt{N}+S}{Sε^2})$, improving the best-known result $O(\frac{N^{2/3}}{S^{2/3}ε^2})$ of SCAFFOLD (Karimireddy et al.,2020) by a factor of $N^{1/6}S^{1/3}$, if the sampled clients $S\leq \sqrt{N}$. Note that in both settings, the communication cost for each round is the same for both FedPAGE and SCAFFOLD. As a result, FedPAGE achieves new state-of-the-art results in terms of communication complexity for both federated convex and nonconvex optimization.

cs.LG↗

arXiv2021

A Short Note of PAGE: Optimal Convergence Rates for Nonconvex Optimization

In this note, we first recall the nonconvex problem setting and introduce the optimal PAGE algorithm (Li et al., ICML'21). Then we provide a simple and clean convergence analysis of PAGE for achieving optimal convergence rates. Moreover, PAGE and its analysis can be easily adopted and generalized to other works. We hope that this note provides the insights and is helpful for future works.

math.OC↗

arXiv2021

PAGE: A Simple and Optimal Probabilistic Gradient Estimator for Nonconvex Optimization

In this paper, we propose a novel stochastic gradient estimator -- ProbAbilistic Gradient Estimator (PAGE) -- for nonconvex optimization. PAGE is easy to implement as it is designed via a small adjustment to vanilla SGD: in each iteration, PAGE uses the vanilla minibatch SGD update with probability $p_t$ or reuses the previous gradient with a small adjustment, at a much lower computational cost, with probability $1-p_t$. We give a simple formula for the optimal choice of $p_t$. Moreover, we prove the first tight lower bound $Ω(n+\frac{\sqrt{n}}{ε^2})$ for nonconvex finite-sum problems, which also leads to a tight lower bound $Ω(b+\frac{\sqrt{b}}{ε^2})$ for nonconvex online problems, where $b:= \min\{\frac{σ^2}{ε^2}, n\}$. Then, we show that PAGE obtains the optimal convergence results $O(n+\frac{\sqrt{n}}{ε^2})$ (finite-sum) and $O(b+\frac{\sqrt{b}}{ε^2})$ (online) matching our lower bounds for both nonconvex finite-sum and online problems. Besides, we also show that for nonconvex functions satisfying the Polyak-Łojasiewicz (PL) condition, PAGE can automatically switch to a faster linear convergence rate $O(\cdot\log \frac{1}ε)$. Finally, we conduct several deep learning experiments (e.g., LeNet, VGG, ResNet) on real datasets in PyTorch showing that PAGE not only converges much faster than SGD in training but also achieves the higher test accuracy, validating the optimal theoretical results and confirming the practical superiority of PAGE.

cs.LG↗

arXiv2020

Acceleration for Compressed Gradient Descent in Distributed and Federated Optimization

Due to the high communication cost in distributed and federated learning problems, methods relying on compression of communicated messages are becoming increasingly popular. While in other contexts the best performing gradient-type methods invariably rely on some form of acceleration/momentum to reduce the number of iterations, there are no methods which combine the benefits of both gradient compression and acceleration. In this paper, we remedy this situation and propose the first accelerated compressed gradient descent (ACGD) methods. In the single machine regime, we prove that ACGD enjoys the rate $O\Big((1+ω)\sqrt{\frac{L}μ}\log \frac{1}ε\Big)$ for $μ$-strongly convex problems and $O\Big((1+ω)\sqrt{\frac{L}ε}\Big)$ for convex problems, respectively, where $ω$ is the compression parameter. Our results improve upon the existing non-accelerated rates $O\Big((1+ω)\frac{L}μ\log \frac{1}ε\Big)$ and $O\Big((1+ω)\frac{L}ε\Big)$, respectively, and recover the optimal rates of accelerated gradient descent as a special case when no compression ($ω=0$) is applied. We further propose a distributed variant of ACGD (called ADIANA) and prove the convergence rate $\widetilde{O}\Big(ω+\sqrt{\frac{L}μ}+\sqrt{\big(\fracω{n}+\sqrt{\fracω{n}}\big)\frac{ωL}μ}\Big)$, where $n$ is the number of devices/workers and $\widetilde{O}$ hides the logarithmic factor $\log \frac{1}ε$. This improves upon the previous best result $\widetilde{O}\Big(ω+ \frac{L}μ+\frac{ωL}{nμ} \Big)$ achieved by the DIANA method of Mishchenko et al. (2019). Finally, we conduct several experiments on real-world datasets which corroborate our theoretical results and confirm the practical superiority of our accelerated methods.

math.OC↗

arXiv2020

A Unified Analysis of Stochastic Gradient Methods for Nonconvex Federated Optimization

In this paper, we study the performance of a large family of SGD variants in the smooth nonconvex regime. To this end, we propose a generic and flexible assumption capable of accurate modeling of the second moment of the stochastic gradient. Our assumption is satisfied by a large number of specific variants of SGD in the literature, including SGD with arbitrary sampling, SGD with compressed gradients, and a wide variety of variance-reduced SGD methods such as SVRG and SAGA. We provide a single convergence analysis for all methods that satisfy the proposed unified assumption, thereby offering a unified understanding of SGD variants in the nonconvex regime instead of relying on dedicated analyses of each variant. Moreover, our unified analysis is accurate enough to recover or improve upon the best-known convergence results of several classical methods, and also gives new convergence results for many new methods which arise as special cases. In the more general distributed/federated nonconvex optimization setup, we propose two new general algorithmic frameworks differing in whether direct gradient compression (DC) or compression of gradient differences (DIANA) is used. We show that all methods captured by these two frameworks also satisfy our unified assumption. Thus, our unified convergence analysis also captures a large variety of distributed methods utilizing compressed communication. Finally, we also provide a unified analysis for obtaining faster linear convergence rates in this nonconvex regime under the PL condition.

math.OC↗

arXiv2020

A Fast Anderson-Chebyshev Acceleration for Nonlinear Optimization

Anderson acceleration (or Anderson mixing) is an efficient acceleration method for fixed point iterations $x_{t+1}=G(x_t)$, e.g., gradient descent can be viewed as iteratively applying the operation $G(x) \triangleq x-α\nabla f(x)$. It is known that Anderson acceleration is quite efficient in practice and can be viewed as an extension of Krylov subspace methods for nonlinear problems. In this paper, we show that Anderson acceleration with Chebyshev polynomial can achieve the optimal convergence rate $O(\sqrtκ\ln\frac{1}ε)$, which improves the previous result $O(κ\ln\frac{1}ε)$ provided by (Toth and Kelley, 2015) for quadratic functions. Moreover, we provide a convergence analysis for minimizing general nonlinear problems. Besides, if the hyperparameters (e.g., the Lipschitz smooth parameter $L$) are not available, we propose a guessing algorithm for guessing them dynamically and also prove a similar convergence rate. Finally, the experimental results demonstrate that the proposed Anderson-Chebyshev acceleration method converges significantly faster than other algorithms, e.g., vanilla gradient descent (GD), Nesterov's Accelerated GD. Also, these algorithms combined with the proposed guessing algorithm (guessing the hyperparameters dynamically) achieve much better performance.

math.OC↗

arXiv2019

A unified variance-reduced accelerated gradient method for convex optimization

We propose a novel randomized incremental gradient algorithm, namely, VAriance-Reduced Accelerated Gradient (Varag), for finite-sum optimization. Equipped with a unified step-size policy that adjusts itself to the value of the condition number, Varag exhibits the unified optimal rates of convergence for solving smooth convex finite-sum problems directly regardless of their strong convexity. Moreover, Varag is the first accelerated randomized incremental gradient method that benefits from the strong convexity of the data-fidelity term to achieve the optimal linear convergence. It also establishes an optimal linear rate of convergence for solving a wide class of problems only satisfying a certain error bound condition rather than strong convexity. Varag can also be extended to solve stochastic finite-sum problems.

math.OC↗

arXiv2019

Gradient Boosting With Piece-Wise Linear Regression Trees

Gradient Boosted Decision Trees (GBDT) is a very successful ensemble learning algorithm widely used across a variety of applications. Recently, several variants of GBDT training algorithms and implementations have been designed and heavily optimized in some very popular open sourced toolkits including XGBoost, LightGBM and CatBoost. In this paper, we show that both the accuracy and efficiency of GBDT can be further enhanced by using more complex base learners. Specifically, we extend gradient boosting to use piecewise linear regression trees (PL Trees), instead of piecewise constant regression trees, as base learners. We show that PL Trees can accelerate convergence of GBDT and improve the accuracy. We also propose some optimization tricks to substantially reduce the training time of PL Trees, with little sacrifice of accuracy. Moreover, we propose several implementation techniques to speedup our algorithm on modern computer architectures with powerful Single Instruction Multiple Data (SIMD) parallelism. The experimental results show that GBDT with PL Trees can provide very competitive testing accuracy with comparable or less training time.

cs.LG↗

arXiv2019

Stochastic Gradient Hamiltonian Monte Carlo with Variance Reduction for Bayesian Inference

Gradient-based Monte Carlo sampling algorithms, like Langevin dynamics and Hamiltonian Monte Carlo, are important methods for Bayesian inference. In large-scale settings, full-gradients are not affordable and thus stochastic gradients evaluated on mini-batches are used as a replacement. In order to reduce the high variance of noisy stochastic gradients, Dubey et al. [2016] applied the standard variance reduction technique on stochastic gradient Langevin dynamics and obtained both theoretical and experimental improvements. In this paper, we apply the variance reduction tricks on Hamiltonian Monte Carlo and achieve better theoretical convergence results compared with the variance-reduced Langevin dynamics. Moreover, we apply the symmetric splitting scheme in our variance-reduced Hamiltonian Monte Carlo algorithms to further improve the theoretical results. The experimental results are also consistent with the theoretical results. As our experiment shows, variance-reduced Hamiltonian Monte Carlo demonstrates better performance than variance-reduced Langevin dynamics in Bayesian regression and classification tasks on real-world datasets.

cs.LG↗

arXiv2019

SSRGD: Simple Stochastic Recursive Gradient Descent for Escaping Saddle Points

We analyze stochastic gradient algorithms for optimizing nonconvex problems. In particular, our goal is to find local minima (second-order stationary points) instead of just finding first-order stationary points which may be some bad unstable saddle points. We show that a simple perturbed version of stochastic recursive gradient descent algorithm (called SSRGD) can find an $(ε,δ)$-second-order stationary point with $\widetilde{O}(\sqrt{n}/ε^2 + \sqrt{n}/δ^4 + n/δ^3)$ stochastic gradient complexity for nonconvex finite-sum problems. As a by-product, SSRGD finds an $ε$-first-order stationary point with $O(n+\sqrt{n}/ε^2)$ stochastic gradients. These results are almost optimal since Fang et al. [2018] provided a lower bound $Ω(\sqrt{n}/ε^2)$ for finding even just an $ε$-first-order stationary point. We emphasize that SSRGD algorithm for finding second-order stationary points is as simple as for finding first-order stationary points just by adding a uniform perturbation sometimes, while all other algorithms for finding second-order stationary points with similar gradient complexity need to combine with a negative-curvature search subroutine (e.g., Neon2 [Allen-Zhu and Li, 2018]). Moreover, the simple SSRGD algorithm gets a simpler analysis. Besides, we also extend our results from nonconvex finite-sum problems to nonconvex online (expectation) problems, and prove the corresponding convergence results.

cs.LG↗

arXiv2019

Stabilized SVRG: Simple Variance Reduction for Nonconvex Optimization

Variance reduction techniques like SVRG provide simple and fast algorithms for optimizing a convex finite-sum objective. For nonconvex objectives, these techniques can also find a first-order stationary point (with small gradient). However, in nonconvex optimization it is often crucial to find a second-order stationary point (with small gradient and almost PSD hessian). In this paper, we show that Stabilized SVRG (a simple variant of SVRG) can find an $ε$-second-order stationary point using only $\widetilde{O}(n^{2/3}/ε^2+n/ε^{1.5})$ stochastic gradients. To our best knowledge, this is the first second-order guarantee for a simple variant of SVRG. The running time almost matches the known guarantees for finding $ε$-first-order stationary points.

cs.LG↗

arXiv2019

Learning Two-layer Neural Networks with Symmetric Inputs

We give a new algorithm for learning a two-layer neural network under a general class of input distributions. Assuming there is a ground-truth two-layer network $$ y = A σ(Wx) + ξ, $$ where $A,W$ are weight matrices, $ξ$ represents noise, and the number of neurons in the hidden layer is no larger than the input or output, our algorithm is guaranteed to recover the parameters $A,W$ of the ground-truth network. The only requirement on the input $x$ is that it is symmetric, which still allows highly complicated and structured input. Our algorithm is based on the method-of-moments framework and extends several results in tensor decompositions. We use spectral algorithms to avoid the complicated non-convex optimization in learning neural networks. Experiments show that our algorithm can robustly learn the ground-truth neural network with a small number of samples for many symmetric input distributions.

cs.LG↗

arXiv2018

A Simple Proximal Stochastic Gradient Method for Nonsmooth Nonconvex Optimization

We analyze stochastic gradient algorithms for optimizing nonconvex, nonsmooth finite-sum problems. In particular, the objective function is given by the summation of a differentiable (possibly nonconvex) component, together with a possibly non-differentiable but convex component. We propose a proximal stochastic gradient algorithm based on variance reduction, called ProxSVRG+. Our main contribution lies in the analysis of ProxSVRG+. It recovers several existing convergence results and improves/generalizes them (in terms of the number of stochastic gradient oracle calls and proximal oracle calls). In particular, ProxSVRG+ generalizes the best results given by the SCSG algorithm, recently proposed by [Lei et al., 2017] for the smooth nonconvex case. ProxSVRG+ is also more straightforward than SCSG and yields simpler analysis. Moreover, ProxSVRG+ outperforms the deterministic proximal gradient descent (ProxGD) for a wide range of minibatch sizes, which partially solves an open problem proposed in [Reddi et al., 2016b]. Also, ProxSVRG+ uses much less proximal oracle calls than ProxSVRG [Reddi et al., 2016b]. Moreover, for nonconvex functions satisfied Polyak-Łojasiewicz condition, we prove that ProxSVRG+ achieves a global linear convergence rate without restart unlike ProxSVRG. Thus, it can \emph{automatically} switch to the faster linear convergence in some regions as long as the objective function satisfies the PL condition locally in these regions. ProxSVRG+ also improves ProxGD and ProxSVRG/SAGA, and generalizes the results of SCSG in this case. Finally, we conduct several experiments and the experimental results are consistent with the theoretical results.

math.OC↗

arXiv2018

Optimal In-Place Suffix Sorting

The suffix array is a fundamental data structure for many applications that involve string searching and data compression. Designing time/space-efficient suffix array construction algorithms has attracted significant attention and considerable advances have been made for the past 20 years. We obtain the \emph{first} in-place suffix array construction algorithms that are optimal both in time and space for (read-only) integer alphabets. Concretely, we make the following contributions: 1. For integer alphabets, we obtain the first suffix sorting algorithm which takes linear time and uses only $O(1)$ workspace (the workspace is the total space needed beyond the input string and the output suffix array). The input string may be modified during the execution of the algorithm, but should be restored upon termination of the algorithm. 2. We strengthen the first result by providing the first in-place linear time algorithm for read-only integer alphabets with $|Σ|=O(n)$ (i.e., the input string cannot be modified). This algorithm settles the open problem posed by Franceschini and Muthukrishnan in ICALP 2007. The open problem asked to design in-place algorithms in $o(n\log n)$ time and ultimately, in $O(n)$ time for (read-only) integer alphabets with $|Σ| \leq n$. Our result is in fact slightly stronger since we allow $|Σ|=O(n)$. 3. Besides, for the read-only general alphabets (i.e., only comparisons are allowed), we present an optimal in-place $O(n\log n)$ time suffix sorting algorithm, recovering the result obtained by Franceschini and Muthukrishnan which was an open problem posed by Manzini and Ferragina in ESA 2002.

cs.DS↗

arXiv2018

A Fast Polynomial-time Primal-Dual Projection Algorithm for Linear Programming

Traditionally, there are several polynomial algorithms for linear programming including the ellipsoid method, the interior point method and other variants. Recently, Chubanov [Chubanov, 2015] proposed a projection and rescaling algorithm, which has become a potentially \emph{practical} class of polynomial algorithms for linear feasibility problems and also for the general linear programming. However, the Chubanov-type algorithms usually perform much better on the infeasible instances than on the feasible instances in practice. To explain this phenomenon, we derive a new theoretical complexity bound for the infeasible instances based on the condition number, which shows that algorithms can indeed run much faster on infeasible instances in certain situations. In order to speed up the feasible instances, we propose a \emph{Polynomial-time Primal-Dual Projection} algorithm (called PPDP) by explicitly developing the dual algorithm. The numerical results validate that our PPDP algorithm achieves a quite balanced performance between feasible and infeasible instances, and its performance is remarkably better than previous algorithms.

math.OC↗

arXiv2018

Modeling and Routing for Predictable Dynamic Networks: This paper is only for copyright protection, and unpublished to the top-level version

The topologies of predictable dynamic networks are continuously dynamic in terms of node position, network connectivity and link metric. However, their dynamics are almost predictable compared with the ad-hoc network. The existing routing protocols specific to static or ad-hoc network do not consider this predictability and thus are not very efficient for some cases. We present a topology model based on Divide-and-Merge methodology to formulate the dynamic topology into the series of static topologies, which can reflect the topology dynamics correctly with the least number of static topologies. Then we design a dynamic programing algorithm to solve that model and determine the timing of routing update and the topology to be used. Besides, for the classic predictable dynamic network---space Internet, the links at some region have shorter delay, which leads to most traffic converge at these links. Meanwhile, the connectivity and metric of these links continuously vary, which results in a great end-to-end path variations and routing updates. In this paper, we propose a stable routing scheme which adds link life-time into its metric to eliminate these dynamics. And then we take use of the Dijkstra's greedy feature to release some paths from the dynamic link, achieving the goal of routing stability. Experimental results show that our method can significantly decrease the number of changed paths and affected network nodes, and then greatly improve the network stability. Interestingly, our method can also achieve better network performance, including the less number of loss packets, smoother variation of end-to-end delay and higher throughput.

cs.NI↗