Towards Federated Learning with On-device Training and Communication in 8-bit Floating Point

Recent work has shown that 8-bit floating point (FP8) can be used for efficiently training neural networks with reduced computational overhead compared to training in FP32/FP16. In this work, we investigate the use of FP8 training in a federated learning context. This brings not only the usual benefits of FP8 which are desirable for on-device training at the edge, but also reduces client-server communication costs due to significant weight compression. We present a novel method for combining FP8 client training while maintaining a global FP32 server model and provide convergence analysis. Experiments with various machine learning models and datasets show that our method consistently yields communication reductions of at least 2.9x across a variety of tasks and models compared to an FP32 baseline.

ALEXR: Optimal Single-Loop Algorithms for Convex Finite-Sum Coupled Compositional Stochastic Optimization

This paper revisits a class of convex Finite-Sum Coupled Compositional Stochastic Optimization (cFCCO) problems with many applications, including group distributionally robust optimization (GDRO), reinforcement learning, and learning to rank. To better solve these problems, we introduce a unified family of efficient single-loop primal-dual block-coordinate proximal algorithms, dubbed ALEXR. This algorithm leverages block-coordinate stochastic mirror ascent updates for the dual variable and stochastic proximal gradient descent updates for the primal variable. We establish the convergence rates of ALEXR in both convex and strongly convex cases under smoothness and non-smoothness conditions of involved functions, which not only improve the best rates in previous works on smooth cFCCO problems but also expand the realm of cFCCO for solving more challenging non-smooth problems such as the dual form of GDRO. Finally, we present lower complexity bounds to demonstrate that the convergence rates of ALEXR are optimal among first-order block-coordinate stochastic algorithms for the considered class of cFCCO problems.

Everything Perturbed All at Once: Enabling Differentiable Graph Attacks

As powerful tools for representation learning on graphs, graph neural networks (GNNs) have played an important role in applications including social networks, recommendation systems, and online web services. However, GNNs have been shown to be vulnerable to adversarial attacks, which can significantly degrade their effectiveness. Recent state-of-the-art approaches in adversarial attacks rely on gradient-based meta-learning to selectively perturb a single edge with the highest attack score until they reach the budget constraint. While effective in identifying vulnerable links, these methods are plagued by high computational costs. By leveraging continuous relaxation and parameterization of the graph structure, we propose a novel attack method called Differentiable Graph Attack (DGA) to efficiently generate effective attacks and meanwhile eliminate the need for costly retraining. Compared to the state-of-the-art, DGA achieves nearly equivalent attack performance with 6 times less training time and 11 times smaller GPU memory footprint on different benchmark datasets. Additionally, we provide extensive experimental analyses of the transferability of the DGA among different graph models, as well as its robustness against widely-used defense mechanisms.

Provable Multi-instance Deep AUC Maximization with Stochastic Pooling

This paper considers a novel application of deep AUC maximization (DAM) for multi-instance learning (MIL), in which a single class label is assigned to a bag of instances (e.g., multiple 2D slices of a CT scan for a patient). We address a neglected yet non-negligible computational challenge of MIL in the context of DAM, i.e., bag size is too large to be loaded into {GPU} memory for backpropagation, which is required by the standard pooling methods of MIL. To tackle this challenge, we propose variance-reduced stochastic pooling methods in the spirit of stochastic optimization by formulating the loss function over the pooled prediction as a multi-level compositional function. By synthesizing techniques from stochastic compositional optimization and non-convex min-max optimization, we propose a unified and provable muli-instance DAM (MIDAM) algorithm with stochastic smoothed-max pooling or stochastic attention-based pooling, which only samples a few instances for each bag to compute a stochastic gradient estimator and to update the model parameter. We establish a similar convergence rate of the proposed MIDAM algorithm as the state-of-the-art DAM algorithms. Our extensive experiments on conventional MIL datasets and medical datasets demonstrate the superiority of our MIDAM algorithm.

GraphFM: Improving Large-Scale GNN Training via Feature Momentum

Training of graph neural networks (GNNs) for large-scale node classification is challenging. A key difficulty lies in obtaining accurate hidden node representations while avoiding the neighborhood explosion problem. Here, we propose a new technique, named feature momentum (FM), that uses a momentum step to incorporate historical embeddings when updating feature representations. We develop two specific algorithms, known as GraphFM-IB and GraphFM-OB, that consider in-batch and out-of-batch data, respectively. GraphFM-IB applies FM to in-batch sampled data, while GraphFM-OB applies FM to out-of-batch data that are 1-hop neighborhood of in-batch data. We provide a convergence analysis for GraphFM-IB and some theoretical insight for GraphFM-OB. Empirically, we observe that GraphFM-IB can effectively alleviate the neighborhood explosion problem of existing methods. In addition, GraphFM-OB achieves promising performance on multiple large-scale graph datasets.

Optimal Algorithms for Stochastic Multi-Level Compositional Optimization

In this paper, we investigate the problem of stochastic multi-level compositional optimization, where the objective function is a composition of multiple smooth but possibly non-convex functions. Existing methods for solving this problem either suffer from sub-optimal sample complexities or need a huge batch size. To address this limitation, we propose a Stochastic Multi-level Variance Reduction method (SMVR), which achieves the optimal sample complexity of $\mathcal{O}\left(1 / \epsilon^{3}\right)$ to find an $\epsilon$-stationary point for non-convex objectives. Furthermore, when the objective function satisfies the convexity or Polyak-{\L}ojasiewicz (PL) condition, we propose a stage-wise variant of SMVR and improve the sample complexity to $\mathcal{O}\left(1 / \epsilon^{2}\right)$ for convex functions or $\mathcal{O}\left(1 /(\mu\epsilon)\right)$ for non-convex functions satisfying the $\mu$-PL condition. The latter result implies the same complexity for $\mu$-strongly convex functions. To make use of adaptive learning rates, we also develop Adaptive SMVR, which achieves the same optimal complexities but converges faster in practice. All our complexities match the lower bounds not only in terms of $\epsilon$ but also in terms of $\mu$ (for PL or strongly convex functions), without using a large batch size in each iteration.

When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee

In this paper, we propose systematic and efficient gradient-based methods for both one-way and two-way partial AUC (pAUC) maximization that are applicable to deep learning. We propose new formulations of pAUC surrogate objectives by using the distributionally robust optimization (DRO) to define the loss for each individual positive data. We consider two formulations of DRO, one of which is based on conditional-value-at-risk (CVaR) that yields a non-smooth but exact estimator for pAUC, and another one is based on a KL divergence regularized DRO that yields an inexact but smooth (soft) estimator for pAUC. For both one-way and two-way pAUC maximization, we propose two algorithms and prove their convergence for optimizing their two formulations, respectively. Experiments demonstrate the effectiveness of the proposed algorithms for pAUC maximization for deep learning on various datasets.

Finite-Sum Coupled Compositional Stochastic Optimization: Theory and Applications

This paper studies stochastic optimization for a sum of compositional functions, where the inner-level function of each summand is coupled with the corresponding summation index. We refer to this family of problems as finite-sum coupled compositional optimization (FCCO). It has broad applications in machine learning for optimizing non-convex or convex compositional measures/objectives such as average precision (AP), $p$-norm push, listwise ranking losses, neighborhood component analysis (NCA), deep survival analysis, deep latent variable models, softmax functions, and model agnostic meta-learning, which deserves finer analysis. Yet, existing algorithms and analysis are restricted in one or other aspects. The contribution of this paper is to provide a comprehensive analysis of a simple stochastic algorithm for both non-convex and convex objectives. The key results are {\bf improved oracle complexities with the parallel speed-up} by the moving-average based stochastic estimator with mini-batching. Our theoretical analysis also exhibits new insights for improving the practical implementation by sampling the batches of equal size for the outer and inner levels. Numerical experiments on AP maximization and $p$-norm push optimization corroborate some aspects of the theory.

Memory-based Optimization Methods for Model-Agnostic Meta-Learning

Recently, model-agnostic meta-learning (MAML) has garnered tremendous attention. However, stochastic optimization of MAML is still immature. Existing algorithms for MAML are based on the ``episode" idea by sampling a number of tasks and a number of data points for each sampled task at each iteration for updating the meta-model. However, they either do not necessarily guarantee convergence with a constant mini-batch size or require processing a larger number of tasks at every iteration, which is not viable for continual learning or cross-device federated learning where only a small number of tasks are available per-iteration or per-round. This paper addresses these issues by (i) proposing efficient memory-based stochastic algorithms for MAML with a diminishing convergence error, which only requires sampling a constant number of tasks and a constant number of examples per-task per-iteration; (ii) proposing communication-efficient distributed memory-based MAML algorithms for personalized federated learning in both the cross-device (w/ client sampling) and the cross-silo (w/o client sampling) settings. The key novelty of the proposed algorithms is to maintain an individual personalized model (aka memory) for each task besides the meta-model and only update them for the sampled tasks by a momentum method that incorporates historical updates at each iteration. The theoretical results significantly improve the optimization theory for MAML and the empirical results also corroborate the theory.

Smoothness-Aware Quantization Techniques

Distributed machine learning has become an indispensable tool for training large supervised machine learning models. To address the high communication costs of distributed training, which is further exacerbated by the fact that modern highly performing models are typically overparameterized, a large body of work has been devoted in recent years to the design of various compression strategies, such as sparsification and quantization, and optimization algorithms capable of using them. Recently, Safaryan et al (2021) pioneered a dramatically different compression design approach: they first use the local training data to form local {\em smoothness matrices}, and then propose to design a compressor capable of exploiting the smoothness information contained therein. While this novel approach leads to substantial savings in communication, it is limited to sparsification as it crucially depends on the linearity of the compression operator. In this work, we resolve this problem by extending their smoothness-aware compression strategy to arbitrary unbiased compression operators, which also includes sparsification. Specializing our results to quantization, we observe significant savings in communication complexity compared to standard quantization. In particular, we show theoretically that block quantization with $n$ blocks outperforms single block quantization, leading to a reduction in communication complexity by an $\mathcal{O}(n)$ factor, where $n$ is the number of nodes in the distributed system. Finally, we provide extensive numerical evidence that our smoothness-aware quantization strategies outperform existing quantization schemes as well the aforementioned smoothness-aware sparsification strategies with respect to all relevant success measures: the number of iterations, the total amount of bits communicated, and wall-clock time.