论文检索

Optimization/Large Scale, Parallel and Distributed

Jiaqi Wang, Xingyi Yang, Suhan Cui, Liwei Che, Lingjuan Lyu, Dongkuan (DK) Xu, Fenglong Ma

This paper focuses on addressing the practical yet challenging problem of model heterogeneity in federated learning, where clients possess models with different network structures. To track this problem, we propose a novel framework called pFedHR, which leverages heterogeneous model reassembly to achieve personalized federated learning. In particular, we approach the problem of heterogeneous model personalization as a model-matching optimization task on the server side. Moreover, pFedHR automatically and dynamically generates informative and diverse personalized candidates with minimal human intervention. Furthermore, our proposed heterogeneous model reassembly technique mitigates the adverse impact introduced by using public data with different distributions from the client data to a certain extent. Experimental results demonstrate that pFedHR outperforms baselines on three datasets under both IID and Non-IID settings. Additionally, pFedHR effectively reduces the adverse impact of using different public data and dynamically generates diverse personalized models in an automated manner.

Optimization/Large Scale, Parallel and Distributed

Yue Tan, Chen Chen, Weiming Zhuang, Xin Dong, Lingjuan Lyu, Guodong Long

Federated learning (FL) is an effective machine learning paradigm where multiple clients can train models based on heterogeneous data in a decentralized manner without accessing their private data. However, existing FL systems undergo performance deterioration due to feature-level test-time shifts, which are well investigated in centralized settings but rarely studied in FL. The common non-IID issue in FL usually refers to inter-client heterogeneity during training phase, while the test-time shift refers to the intra-client heterogeneity during test phase. Although the former is always deemed to be notorious for FL, there is still a wealth of useful information delivered by heterogeneous data sources, which may potentially help alleviate the latter issue. To explore the possibility of using inter-client heterogeneity in handling intra-client heterogeneity, we firstly propose a contrastive learning-based FL framework, namely FedICON, to capture invariant knowledge among heterogeneous clients and consistently tune the model to adapt to test data. In FedICON, each client performs sample-wise supervised contrastive learning during the local training phase, which enhances sample-wise invariance encoding ability. Through global aggregation, the invariance extraction ability can be mutually boosted among inter-client heterogeneity. During the test phase, our test-time adaptation procedure leverages unsupervised contrastive learning to guide the model to smoothly generalize to test data under intra-client heterogeneity. Extensive experiments validate the effectiveness of the proposed FedICON in taming heterogeneity to handle test-time shift problems.

Optimization/Large Scale, Parallel and Distributed

Jingfeng Wu, Wennan Zhu, Peter Kairouz, Vladimir Braverman

In federated frequency estimation (FFE), multiple clients work together to estimate the frequency of their local data by communicating with a server, while maintaining the security constraint of $\mathtt{secsum}$ where the server can only access the sum of client-held vectors. For FFE with a single communication round, it is known that count sketch is nearly information-theoretically optimal [Chen et al., 2022]. However, when multiple communication rounds are allowed, we propose a new sketch algorithm that is provably more accurate than a naive adaptation of count sketch. Furthermore, we show that both our sketch algorithm and count sketch can achieve better accuracy when the problem instance is simpler. Therefore, we propose a two-phase approach to enable the use of a smaller sketch size for simpler problems. Finally, we provide mechanisms to make our proposed algorithm differentially private. We verify the performance of our methods through experiments conducted on real datasets.

Optimization/Large Scale, Parallel and Distributed

Martijn De Vos, Sadegh Farhadkhani, Rachid Guerraoui, Anne-marie Kermarrec, Rafael Pires, Rishi Sharma

We present Epidemic Learning (EL), a simple yet powerful decentralized learning (DL) algorithm that leverages changing communication topologies to achieve faster model convergence compared to conventional DL approaches. At each round of EL, each node sends its model updates to a random sample of $s$ other nodes (in a system of $n$ nodes). We provide an extensive theoretical analysis of EL, demonstrating that its changing topology culminates in superior convergence properties compared to the state-of-the-art (static and dynamic) topologies. Considering smooth non-convex loss functions, the number of transient iterations for EL, i.e., the rounds required to achieve asymptotic linear speedup, is in $O(n^3/s^2)$ which outperforms the best-known bound $O(n^3)$ by a factor of $s^2$, indicating the benefit of randomized communication for DL. We empirically evaluate EL in a 96-node network and compare its performance with state-of-the-art DL approaches. Our results illustrate that EL converges up to $ 1.7\times$ quicker than baseline DL algorithms and attains $2.2 $\% higher accuracy for the same communication volume.

Optimization/Large Scale, Parallel and Distributed

Guanqiang Zhou, Ping Xu, Yue Wang, Zhi Tian

Fairness and robustness are two important goals in the design of modern distributed learning systems. Despite a few prior works attempting to achieve both fairness and robustness, some key aspects of this direction remain underexplored. In this paper, we try to answer three largely unnoticed and unaddressed questions that are of paramount significance to this topic: (i) What makes jointly satisfying fairness and robustness difficult? (ii) Is it possible to establish theoretical guarantee for the dual property of fairness and robustness? (iii) How much does fairness have to sacrifice at the expense of robustness being incorporated into the system? To address these questions, we first identify data heterogeneity as the key difficulty of combining fairness and robustness. Accordingly, we propose a fair and robust framework called H-nobs which can offer certified fairness and robustness through the adoption of two key components, a fairness-promoting objective function and a simple robust aggregation scheme called norm-based screening (NBS). We explain in detail why NBS is the suitable scheme in our algorithm in contrast to other robust aggregation measures. In addition, we derive three convergence theorems for H-nobs in cases of the learning model being nonconvex, convex, and strongly convex respectively, which provide theoretical guarantees for both fairness and robustness. Further, we empirically investigate the influence of the robust mechanism (NBS) on the fairness performance of H-nobs, the very first attempt of such exploration.

Optimization/Large Scale, Parallel and Distributed

Sai Aparna Aketi, Abolfazl Hashemi, Kaushik Roy

Decentralized learning enables the training of deep learning models over large distributed datasets generated at different locations, without the need for a central server. However, in practical scenarios, the data distribution across these devices can be significantly different, leading to a degradation in model performance. In this paper, we focus on designing a decentralized learning algorithm that is less susceptible to variations in data distribution across devices. We propose Global Update Tracking (GUT), a novel tracking-based method that aims to mitigate the impact of heterogeneous data in decentralized learning without introducing any communication overhead. We demonstrate the effectiveness of the proposed technique through an exhaustive set of experiments on various Computer Vision datasets (CIFAR-10, CIFAR-100, Fashion MNIST, and ImageNette), model architectures, and network topologies. Our experiments show that the proposed method achieves state-of-the-art performance for decentralized learning on heterogeneous data via a 1-6% improvement in test accuracy compared to other existing techniques.

Optimization/Large Scale, Parallel and Distributed

Zachary Charles, Nicole Mitchell, Krishna Pillutla, Michael Reneer, Zachary Garrett

We introduce Dataset Grouper, a library to create large-scale group-structured (e.g., federated) datasets, enabling federated learning simulation at the scale of foundation models. This library facilitates the creation of group-structured versions of existing datasets based on user-specified partitions, and directly leads to a variety of useful heterogeneous datasets that can be plugged into existing software frameworks. Dataset Grouper offers three key advantages. First, it scales to settings where even a single group's dataset is too large to fit in memory. Second, it provides flexibility, both in choosing the base (non-partitioned) dataset and in defining partitions. Finally, it is framework-agnostic. We empirically demonstrate that Dataset Grouper enables large-scale federated language modeling simulations on datasets that are orders of magnitude larger than in previous work, allowing for federated training of language models with hundreds of millions, and even billions, of parameters. Our experimental results show that algorithms like FedAvg operate more as meta-learning methods than as empirical risk minimization methods at this scale, suggesting their utility in downstream personalization and task-specific adaptation. Dataset Grouper is available at https://github.com/google-research/dataset_grouper.

Optimization/Large Scale, Parallel and Distributed

Alexander Tyurin, Peter Richtarik

We present a new method that includes three key components of distributed optimization and federated learning: variance reduction of stochastic gradients, partial participation, and compressed communication. We prove that the new method has optimal oracle complexity and state-of-the-art communication complexity in the partial participation setting. Regardless of the communication compression feature, our method successfully combines variance reduction and partial participation: we get the optimal oracle complexity, never need the participation of all nodes, and do not require the bounded gradients (dissimilarity) assumption.

Optimization/Large Scale, Parallel and Distributed

Alexander Tyurin, Peter Richtarik

We consider distributed convex optimization problems in the regime when the communication between the server and the workers is expensive in both uplink and downlink directions. We develop a new and provably accelerated method, which we call 2Direction, based on fast bidirectional compressed communication and a new bespoke error-feedback mechanism which may be of independent interest. Indeed, we find that the EF and EF21-P mechanisms (Seide et al., 2014; Gruntkowska et al., 2023) that have considerable success in the design of efficient non-accelerated methods are not appropriate for accelerated methods. In particular, we prove that 2Direction improves the previous state-of-the-art communication complexity $\widetilde{\Theta}\left(K \times \left(\frac{L}{\alpha \mu} + \frac{L_{\max} \omega}{n \mu} + \omega\right)\right)$ (Gruntkowska et al., 2023) to $\widetilde{\Theta}(K \times (\sqrt{\frac{L (\omega + 1)}{\alpha \mu}} + \sqrt{\frac{L_{\max} \omega^2}{n \mu}} + \frac{1}{\alpha} + \omega))$ in the $\mu$--strongly-convex setting, where $L$ and $L_{\max}$ are smoothness constants, $n$ is \# of workers, $\omega$ and $\alpha$ are compression errors of the Rand$K$ and Top$K$ sparsifiers (as examples), $K$ is \# of coordinates/bits that the server and workers send to each other. Moreover, our method is the first that improves upon the communication complexity of the vanilla accelerated gradient descent method (AGD). We obtain similar improvements in the general convex regime as well. Finally, our theoretical findings are corroborated by experimental evidence.

Optimization/Large Scale, Parallel and Distributed

Alexander Tyurin, Peter Richtarik

Parallelization is a popular strategy for improving the performance of methods. Optimization methods are no exception: design of efficient parallel optimization methods and tight analysis of their theoretical properties are important research endeavors. While the minimax complexities are well known for sequential optimization methods, the theory of parallel optimization methods is less explored. In this paper, we propose a new protocol that generalizes the classical oracle framework approach. Using this protocol, we establish minimax complexities for parallel optimization methods that have access to an unbiased stochastic gradient oracle with bounded variance. We consider a fixed computation model characterized by each worker requiring a fixed but worker-dependent time to calculate stochastic gradient. We prove lower bounds and develop optimal algorithms that attain them. Our results have surprising consequences for the literature of asynchronous optimization methods.

Optimization/Large Scale, Parallel and Distributed

Yuki Takezawa, Ryoma Sato, Han Bao, Kenta Niwa, Makoto Yamada

Decentralized learning has recently been attracting increasing attention for its applications in parallel computation and privacy preservation. Many recent studies stated that the underlying network topology with a faster consensus rate (a.k.a. spectral gap) leads to a better convergence rate and accuracy for decentralized learning. However, a topology with a fast consensus rate, e.g., the exponential graph, generally has a large maximum degree, which incurs significant communication costs. Thus, seeking topologies with both a fast consensus rate and small maximum degree is important. In this study, we propose a novel topology combining both a fast consensus rate and small maximum degree called the Base-$\left(k+1\right)$ Graph. Unlike the existing topologies, the Base-$\left(k+1\right)$ Graph enables all nodes to reach the exact consensus after a finite number of iterations for any number of nodes and maximum degree $k$. Thanks to this favorable property, the Base-$\left(k+1\right)$ Graph endows Decentralized SGD (DSGD) with both a faster convergence rate and more communication efficiency than the exponential graph. We conducted experiments with various topologies, demonstrating that the Base-$\left(k+1\right)$ Graph enables various decentralized learning methods to achieve higher accuracy with better communication efficiency than the existing topologies. Our code is available at https://github.com/yukiTakezawa/BaseGraph.

Optimization/Large Scale, Parallel and Distributed

Ilyas Fatkhullin, Alexander Tyurin, Peter Richtarik

Due to the high communication overhead when training machine learning models in a distributed environment, modern algorithms invariably rely on lossy communication compression. However, when untreated, the errors caused by compression propagate, and can lead to severely unstable behavior, including exponential divergence. Almost a decade ago, Seide et al. [2014] proposed an error feedback (EF) mechanism, which we refer to as EF14, as an immensely effective heuristic for mitigating this issue. However, despite steady algorithmic and theoretical advances in the EF field in the last decade, our understanding is far from complete. In this work we address one of the most pressing issues. In particular, in the canonical nonconvex setting, all known variants of EF rely on very large batch sizes to converge, which can be prohibitive in practice. We propose a surprisingly simple fix which removes this issue both theoretically, and in practice: the application of Polyak's momentum to the latest incarnation of EF due to Richtárik et al. [2021] known as EF21. Our algorithm, for which we coin the name EF21-SGDM, improves the communication and sample complexities of previous error feedback algorithms under standard smoothness and bounded variance assumptions, and does not require any further strong assumptions such as bounded gradient dissimilarity. Moreover, we propose a double momentum version of our method that improves the complexities even further. Our proof seems to be novel even when compression is removed form the method, and as such, our proof technique is of independent interest in the study of nonconvex stochastic optimization enriched with Polyak's momentum.

Optimization/Large Scale, Parallel and Distributed

Tahseen Rabbani, Marco Bornstein, Furong Huang

Locality-sensitive hashing (LSH) based frameworks have been used efficiently to select weight vectors in a dense hidden layer with high cosine similarity to an input, enabling dynamic pruning. While this type of scheme has been shown to improve computational training efficiency, existing algorithms require repeated randomized projection of the full layer weight, which is impractical for computational- and memory-constrained devices. In a distributed setting, deferring LSH analysis to a centralized host is (i) slow if the device cluster is large and (ii) requires access to input data which is forbidden in a federated context. Using a new family of hash functions, we develop the first private, personalized, and memory-efficient on-device LSH framework.Our framework enables privacy and personalization by allowing each device to generate hash tables, without the help of a central host, using device-specific hashing hyper-parameters (e.g., number of hash tables or hash length).Hash tables are generated with a compressed set of the full weights, and can be serially generated and discarded if the process is memory-intensive.This allows devices to avoid maintaining (i) the fully-sized model and (ii) large amounts of hash tables in local memory for LSH analysis. We prove several statistical and sensitivity properties of our hash functions, and experimentally demonstrate that our framework is competitive in training large scale recommender networks compared to other LSH frameworks which assume unrestricted on-device capacity.

Optimization/Large Scale, Parallel and Distributed

Kilian Pfeiffer, Ramin Khalili, Joerg Henkel

Federated learning (FL) is usually performed on resource-constrained edge devices, e.g., with limited memory for the computation. If the required memory to train a model exceeds this limit, the device will be excluded from the training. This can lead to a lower accuracy as valuable data and computation resources are excluded from training, also causing bias and unfairness. The FL training process should be adjusted to such constraints. The state-of-the-art techniques propose training subsets of the FL model at constrained devices, reducing their resource requirements for training. However, these techniques largely limit the co-adaptation among parameters of the model and are highly inefficient, as we show: it is actually better to train a smaller (less accurate) model by the system where all the devices can train the model end-to-end than applying such techniques. We propose a new method that enables successive freezing and training of the parameters of the FL model at devices, reducing the training’s resource requirements at the devices while still allowing enough co-adaptation between parameters. We show through extensive experimental evaluation that our technique greatly improves the accuracy of the trained model (by 52.4 p.p. ) compared with the state of the art, efficiently aggregating the computation capacity available on distributed devices.

Optimization/Large Scale, Parallel and Distributed

Wenzhi Gao, Qi Deng

This paper studies delayed stochastic algorithms for weakly convex optimization in a distributed network with workers connected to a master node. Recently, Xu~et~al.~2022 showed that an inertial stochastic subgradient method converges at a rate of $\mathcal{O}(\tau_{\text{max}}/\sqrt{K})$ which depends on the maximum information delay $\tau_{\text{max}}$. In this work, we show that the delayed stochastic subgradient method ($\texttt{DSGD}$) obtains a tighter convergence rate which depends on the expected delay $\bar{\tau}$. Furthermore, for an important class of composition weakly convex problems, we develop a new delayed stochastic prox-linear ($\texttt{DSPL}$) method in which the delays only affect the high-order term in the rate and hence, are negligible after a certain number of $\texttt{DSPL}$ iterations. In addition, we demonstrate the robustness of our proposed algorithms against arbitrary delays. By incorporating a simple safeguarding step in both methods, we achieve convergence rates that depend solely on the number of workers, eliminating the effect of delays. Our numerical experiments further confirm the empirical superiority of our proposed methods.

Optimization/Large Scale, Parallel and Distributed

Irene Wang, Prashant Nair, Divya Mahajan

Federated Learning (FL) allows machine learning models to train locally on individual mobile devices, synchronizing model updates via a shared server. This approach safeguards user privacy; however, it also generates a heterogeneous training environment due to the varying performance capabilities across devices. As a result, “straggler” devices with lower performance often dictate the overalltraining time in FL. In this work, we aim to alleviate this performance bottleneck due to stragglers by dynamically balancing the training load across the system. We introduce Invariant Dropout, a method that extracts a sub-model based on the weight update threshold, thereby minimizing potential impacts on accuracy. Building on this dropout technique, we develop an adaptive training framework, Federated Learning using Invariant Dropout (FLuID). FLuID offers a lightweight sub-model extraction to regulate computational intensity, thereby reducing the load on straggler devices without affecting model quality. Our method leverages neuron updates from non-straggler devices to construct a tailored sub-model for each straggler based on client performance profiling. Furthermore, FLuID can dynamically adapt to changes in stragglers as runtime conditions shift. We evaluate FLuID using five real-world mobile clients. The evaluations show that Invariant Dropout maintains baseline model efficiency while alleviating the performance bottleneck of stragglers through a dynamic, runtime approach.

Optimization/Large Scale, Parallel and Distributed

Shaochen (Henry) Zhong, Zaichuan You, Jiamu Zhang, Sebastian Zhao, Zachary LeClaire, Zirui Liu, Daochen Zha, Vipin Chaudhary, Shuai Xu, Xia Hu

Densely structured pruning methods utilizing simple pruning heuristics can deliver immediate compression and acceleration benefits with acceptable benign performances. However, empirical findings indicate such naively pruned networks are extremely fragile under simple adversarial attacks. Naturally, we would be interested in knowing if such a phenomenon also holds for carefully designed modern structured pruning methods. If so, then to what extent is the severity? And what kind of remedies are available? Unfortunately, both the questions and the solution remain largely unaddressed: no prior art is able to provide a thorough investigation on the adversarial performance of modern structured pruning methods (spoiler: it is not good), yet the few works that attempt to provide mitigation often do so at various extra costs with only to-be-desired performance.In this work, we answer both questions by fairly and comprehensively investigating the adversarial performance of 10+ popular structured pruning methods. Solution-wise, we take advantage of *Grouped Kernel Pruning (GKP)*'s recent success in pushing densely structured pruning freedom to a more fine-grained level. By mixing up kernel smoothness — a classic robustness-related kernel-level metric — into a modified GKP procedure, we present a one-shot-post-train-weight-dependent GKP method capable of advancing SOTA performance on both the benign and adversarial scale, while requiring no extra (in fact, often less) cost than a standard pruning procedure. Please refer to our [GitHub repository](https://github.com/henryzhongsc/adv_robust_gkp) for code implementation, tool sharing, and model checkpoints.

Optimization/Large Scale, Parallel and Distributed

Niv Giladi, Shahar Gottlieb, moran shkolnik, Asaf Karnieli, Ron Banner, Elad Hoffer, Kfir Y. Levy, Daniel Soudry

Background: Distributed training is essential for large scale training of deep neural networks (DNNs). The dominant methods for large scale DNN training are synchronous (e.g. All-Reduce), but these require waiting for all workers in each step. Thus, these methods are limited by the delays caused by straggling workers.Results: We study a typical scenario in which workers are straggling due to variability in compute time. We find an analytical relation between compute time properties and scalability limitations, caused by such straggling workers. With these findings, we propose a simple yet effective decentralized method to reduce the variation among workers and thus improve the robustness of synchronous training. This method can be integrated with the widely used All-Reduce. Our findings are validated on large-scale training tasks using 200 Gaudi Accelerators.

Optimization/Large Scale, Parallel and Distributed

Sepideh Mahabadi, Stojan Trajanovski

We study core-set construction algorithms for the task of Diversity Maximization under fairness/partition constraint. Given a set of points $P$ in a metric space partitioned into $m$ groups, and given $k_1,\ldots,k_m$, the goal of this problem is to pick $k_i$ points from each group $i$ such that the overall diversity of the $k=\sum_i k_i$ picked points is maximized. We consider two natural diversity measures: sum-of-pairwise distances and sum-of-nearest-neighbor distances, and show improved core-set construction algorithms with respect to these measures. More precisely, we show the first constant factor core-set w.r.t. sum-of-pairwise distances whose size is independent of the size of the dataset and the aspect ratio. Second, we show the first core-set w.r.t. the sum-of-nearest-neighbor distances. Finally, we run several experiments showing the effectiveness of our core-set approach. In particular, we apply constrained diversity maximization to summarize a set of timed messages that takes into account the messages' recency. Specifically, the summary should include more recent messages compared to older ones. This is a real task in one of the largest communication platforms, affecting the experience of hundreds of millions daily active users. By utilizing our core-set method for this task, we achieve a 100x speed-up while losing the diversity by only a few percent. Moreover, our approach allows us to improve the space usage of the algorithm in the streaming setting.

Optimization/Large Scale, Parallel and Distributed

Yuzhong Wang, Xu Han, Weilin Zhao, Guoyang Zeng, Zhiyuan Liu, Maosong Sun

In recent years, big models based on Transformers have achieved state-of-the-art performance on many artificial intelligence (AI) tasks.Despite the success of these Transformer-based models, their huge parameter size poses a serious challenge to their training, both from the storage and computation perspectives.To this end, memory optimization (e.g., rematerialization and offloading) and parallelism (e.g., data parallelism and model parallelism) are widely explored to make training Transformers more efficient.In this paper, we propose a framework to automatically find an efficient integration of memory optimization and parallelism for High-Throughput Transformer Training (named H3T), which is rarely considered by existing efforts for training big Transformer-based models.Specifically, we design search algorithms to combine appropriate memory optimization strategies and parallelism schemes to achieve a balance between memory overhead and training efficiency.We implement H3T based on an open-source toolkit BMTrain and then use H3T to train the Transformers of different sizes to evaluate the efficiency of H3T.The experimental results show that H3T outperforms the most popular deep learning (DL) toolkit Megatron-DeepSpeed by $1.2\times \sim 4.3\times$ training speed while reducing $34.6\% \sim 80.5\%$ of memory overhead.Moreover, H3T can use only 64 NVIDIA A100 GPUs to train GPT-3-175B, which is very difficult for existing DL toolkits. The source code is available at https://github.com/OpenBMB/BMTrain/tree/h3t.