Running Stochastic Gradient Descent (SGD) in a decentralized fashion has shown promising results. In this paper we propose Moniqua, a technique that allows decentralized SGD to use quantized communication. We prove in theory that Moniqua communicates a provably bounded number of bits per iteration, while converging at the same asymptotic rate as the original algorithm does with full-precision communication. Moniqua improves upon prior works in that it (1) requires zero additional memory, (2) works with 1-bit quantization, and (3) is applicable to a variety of decentralized algorithms. We demonstrate empirically that Moniqua converges faster with respect to wall clock time than other quantized decentralized algorithms. We also show that Moniqua is robust to very low bit-budgets, allowing 1-bit-per-parameter communication without compromising validation accuracy when training ResNet20 and ResNet110 on CIFAR10.

The existence of a saddle point in convex-concave games follows from the celebrated minimax theorem (von Neumann, 1928; Sion et al., 1958) and numerical algorithms for finding it has a long history in optimization (Dem'yanov and Malozemov, 1974; Nemirovsky and Yudin, 1983; Zhang et al., 2019; Lin et al., 2020). Recent success in generative adversarial networks (GANs) (Goodfellow et al., 2014; Heusel et al., 2017) and reinforcement learning (Sutton et al., 1998; Du et al., 2017; Dai et al., 2018) has imposed new challenges for non-convex-non-concave (NCNC) settings. An important question is: what is a local optimal point in min-max optimization? Daskalakis and Panageas (2018) used a local version of saddle points to define local optimality and studied the local convergence behavior of gradient descent (GD) (Arrow et al., 1958) and optimistic gradient descent (OGD) (Popov, 1980; Daskalakis et al., 2018). Following this work, Jin et al. (2019) proposed a new definition of local optimality called local min-max points, and studied how GD converges to such points. One of the difficulties in the NCNC settings is the absence of strong duality (e.g. Boyd and Vandenberghe (2004)), which implies an intrinsic order for min-max optimization, known as Stackelberg games (von Stackelberg, 1934) where a leader takes action first and a follower acts upon the leader's strategy. The global solution to Stackelberg games is well-defined, which we call a global min-max point (a.k.a.

The ubiquitous use of machine learning algorithms brings new challenges to traditional database problems such as incremental view update. Much effort is being put in better understanding and debugging machine learning models, as well as in identifying and repairing errors in training datasets. Our focus is on how to assist these activities when they have to retrain the machine learning model after removing problematic training samples in cleaning or selecting different subsets of training data for interpretability. This paper presents an efficient provenance-based approach, PrIU, and its optimized version, PrIU-opt, for incrementally updating model parameters without sacrificing prediction accuracy. We prove the correctness and convergence of the incrementally updated model parameters, and validate it experimentally. Experimental results show that up to two orders of magnitude speed-ups can be achieved by PrIU-opt compared to simply retraining the model from scratch, yet obtaining highly similar models.

Existing research shows that the batch size can seriously affect the performance of stochastic gradient descent~(SGD) based learning, including training speed and generalization ability. A larger batch size typically results in less parameter updates. In distributed training, a larger batch size also results in less frequent communication. However, a larger batch size can make a generalization gap more easily. Hence, how to set a proper batch size for SGD has recently attracted much attention. Although some methods about setting batch size have been proposed, the batch size problem has still not been well solved. In this paper, we first provide theory to show that a proper batch size is related to the gap between initialization and optimum of the model parameter. Then based on this theory, we propose a novel method, called \underline{s}tagewise \underline{e}nlargement of \underline{b}atch \underline{s}ize~(\mbox{SEBS}), to set proper batch size for SGD. More specifically, \mbox{SEBS} adopts a multi-stage scheme, and enlarges the batch size geometrically by stage. We theoretically prove that, compared to classical stagewise SGD which decreases learning rate by stage, \mbox{SEBS} can reduce the number of parameter updates without increasing generalization error. SEBS is suitable for \mbox{SGD}, momentum \mbox{SGD} and AdaGrad. Empirical results on real data successfully verify the theories of \mbox{SEBS}. Furthermore, empirical results also show that SEBS can outperform other baselines.

We consider the problem of optimizing an objective function with and without convexity in a simulation-optimization context, where only stochastic zeroth-order information is available. We consider two techniques for estimating gradient/Hessian, namely simultaneous perturbation (SP) and Gaussian smoothing (GS). We introduce an optimization oracle to capture a setting where the function measurements have an estimation error that can be controlled. Our oracle is appealing in several practical contexts where the objective has to be estimated from i.i.d. samples, and increasing the number of samples reduces the estimation error. In the stochastic non-convex optimization context, we analyze the zeroth-order variant of the randomized stochastic gradient (RSG) and quasi-Newton (RSQN) algorithms with a biased gradient/Hessian oracle, and with its variant involving an estimation error component. In particular, we provide non-asymptotic bounds on the performance of both algorithms, and our results provide a guideline for choosing the batch size for estimation, so that the overall error bound matches with the one obtained when there is no estimation error. Next, in the stochastic convex optimization setting, we provide non-asymptotic bounds that hold in expectation for the last iterate of a stochastic gradient descent (SGD) algorithm, and our bound for the GS variant of SGD matches the bound for SGD with unbiased gradient information. We perform simulation experiments on synthetic as well as real-world datasets, and the empirical results validate the theoretical findings.

Deep neural networks (DNNs) for supervised learning can be viewed as a pipeline of a feature extractor (i.e. last hidden layer) and a linear classifier (i.e. output layer) that is trained jointly with stochastic gradient descent (SGD). In each iteration of SGD, a mini-batch from the training data is sampled and the true gradient of the loss function is estimated as the noisy gradient calculated on this mini-batch. From the feature learning perspective, the feature extractor should be updated to learn meaningful features with respect to the entire data, and reduce the accommodation to noise in the mini-batch. With this motivation, we propose In-Training Distribution Matching (ITDM) to improve DNN training and reduce overfitting. Specifically, along with the loss function, ITDM regularizes the feature extractor by matching the moments of distributions of different mini-batches in each iteration of SGD, which is fulfilled by minimizing the maximum mean discrepancy. As such, ITDM does not assume any explicit parametric form of data distribution in the latent feature space. Extensive experiments are conducted to demonstrate the effectiveness of our proposed strategy.

We consider the problem of meta-learning, or learning-to-learn [Thrun and Pratt, 1998], in which the goal is to use the data from numerous training tasks to reduce the sample complexity of an unseen but related test task. Although there is a long history of successful methods in meta-learning and the related areas of multi-task and lifelong learning [Evgeniou and Pontil, 2004, Ruvolo and Eaton, 2013], recent approaches have been developed with the diversity and scale of modern applications in mind. This has given rise to simple, model-agnostic methods that focus on learning a good initialization for some gradient-based method such as stochastic gradient descent (SGD), to be run on samples from a new task [Finn et al., 2017, Nichol et al., 2018]. These methods have found widespread applications in a variety of areas such as computer vision [Nichol et al., 2018], reinforcement learning [Finn et al., 2017], and federated learning [McMahan et al., 2017]. Inspired by their popularity, several recent learning-theoretic analyses of meta-learning have followed suit, eschewing customization to specific hypothesis classes such as halfspaces [Maurer and Pontil, 2013, Balcan et al., 2015] and instead favoring the convex-case study of gradient-based algorithms that could potentially be applied to deep neural networks [Denevi et al., 2019, Khodak et al., 2019].

We consider the setting where a master wants to run a distributed stochastic gradient descent (SGD) algorithm on $n$ workers each having a subset of the data. Distributed SGD may suffer from the effect of stragglers, i.e., slow or unresponsive workers who cause delays. One solution studied in the literature is to wait at each iteration for the responses of the fastest $k

Graph Neural Networks (GNNs) have made significant advances on several fundamental inference tasks. As a result, there is a surge of interest in using these models for making potentially important decisions in high-regret applications. However, despite GNNs' impressive performance, it has been observed that carefully crafted perturbations on graph structures (or nodes attributes) lead them to make wrong predictions. Presence of these adversarial examples raises serious security concerns. Most of the existing robust GNN design/training methods are only applicable to white-box settings where model parameters are known and gradient based methods can be used by performing convex relaxation of the discrete graph domain. More importantly, these methods are not efficient and scalable which make them infeasible in time sensitive tasks and massive graph datasets. To overcome these limitations, we propose a general framework which leverages the greedy search algorithms and zeroth-order methods to obtain robust GNNs in a generic and an efficient manner. On several applications, we show that the proposed techniques are significantly less computationally expensive and, in some cases, more robust than the state-of-the-art methods making them suitable to large-scale problems which were out of the reach of traditional robust training methods.

Many binary classification problems minimize misclassification above (or below) a threshold. We show that instances of ranking problems, accuracy at the top or hypothesis testing may be written in this form. We propose a general framework to handle these classes of problems and show which known methods (both known and newly proposed) fall into this framework. We provide a theoretical analysis of this framework and mention selected possible pitfalls the methods may encounter. We suggest several numerical improvements including the implicit derivative and stochastic gradient descent. We provide an extensive numerical study. Based both on the theoretical properties and numerical experiments, we conclude the paper by suggesting which method should be used in which situation.