Abdellatif Zaidi

Milad Sefidgaran, Romain Chor, Abdellatif Zaidi

In this paper, we use tools from rate-distortion theory to establish new upper bounds on the generalization error of statistical distributed learning algorithms. Specifically, there are $K$ clients whose individually chosen models are aggregated by a central server. The bounds depend on the compressibility of each client's algorithm while keeping other clients' algorithms un-compressed, and leverage the fact that small changes in each local model change the aggregated model by a factor of only $1/K$. Adopting a recently proposed approach by Sefidgaran et al., and extending it suitably to the distributed setting, this enables smaller rate-distortion terms which are shown to translate into tighter generalization bounds. The bounds are then applied to the distributed support vector machines (SVM), suggesting that the generalization error of the distributed setting decays faster than that of the centralized one with a factor of $\mathcal{O}(\log(K)/\sqrt{K})$. This finding is validated also experimentally. A similar conclusion is obtained for a multiple-round federated learning setup where each client uses stochastic gradient Langevin dynamics (SGLD).

Yijun Wan, Melih Barsbey, Abdellatif Zaidi et al.

Neural network compression has been an increasingly important subject, not only due to its practical relevance, but also due to its theoretical implications, as there is an explicit connection between compressibility and generalization error. Recent studies have shown that the choice of the hyperparameters of stochastic gradient descent (SGD) can have an effect on the compressibility of the learned parameter vector. These results, however, rely on unverifiable assumptions and the resulting theory does not provide a practical guideline due to its implicitness. In this study, we propose a simple modification for SGD, such that the outputs of the algorithm will be provably compressible without making any nontrivial assumptions. We consider a one-hidden-layer neural network trained with SGD, and show that if we inject additive heavy-tailed noise to the iterates at each iteration, for any compression rate, there exists a level of overparametrization such that the output of the algorithm will be compressible with high probability. To achieve this result, we make two main technical contributions: (i) we prove a 'propagation of chaos' result for a class of heavy-tailed stochastic differential equations, and (ii) we derive error estimates for their Euler discretization. Our experiments suggest that the proposed approach not only achieves increased compressibility with various models and datasets, but also leads to robust test performance under pruning, even in more realistic architectures that lie beyond our theoretical setting.

Romain Chor, Milad Sefidgaran, Abdellatif Zaidi

We study the generalization error of statistical learning models in a Federated Learning (FL) setting. Specifically, there are $K$ devices or clients, each holding an independent own dataset of size $n$. Individual models, learned locally via Stochastic Gradient Descent, are aggregated (averaged) by a central server into a global model and then sent back to the devices. We consider multiple (say $R \in \mathbb N^*$) rounds of model aggregation and study the effect of $R$ on the generalization error of the final aggregated model. We establish an upper bound on the generalization error that accounts explicitly for the effect of $R$ (in addition to the number of participating devices $K$ and dataset size $n$). It is observed that, for fixed $(n, K)$, the bound increases with $R$, suggesting that the generalization of such learning algorithms is negatively affected by more frequent communication with the parameter server. Combined with the fact that the empirical risk, however, generally decreases for larger values of $R$, this indicates that $R$ might be a parameter to optimize to reduce the population risk of FL algorithms. The results of this paper, which extend straightforwardly to the heterogeneous data setting, are also illustrated through numerical examples.

Milad Sefidgaran, Abdellatif Zaidi

In this paper, we establish novel data-dependent upper bounds on the generalization error through the lens of a "variable-size compressibility" framework that we introduce newly here. In this framework, the generalization error of an algorithm is linked to a variable-size 'compression rate' of its input data. This is shown to yield bounds that depend on the empirical measure of the given input data at hand, rather than its unknown distribution. Our new generalization bounds that we establish are tail bounds, tail bounds on the expectation, and in-expectations bounds. Moreover, it is shown that our framework also allows to derive general bounds on any function of the input data and output hypothesis random variables. In particular, these general bounds are shown to subsume and possibly improve over several existing PAC-Bayes and data-dependent intrinsic dimension-based bounds that are recovered as special cases, thus unveiling a unifying character of our approach. For instance, a new data-dependent intrinsic dimension-based bound is established, which connects the generalization error to the optimization trajectories and reveals various interesting connections with the rate-distortion dimension of a process, the Rényi information dimension of a process, and the metric mean dimension.

Milad Sefidgaran, Romain Chor, Abdellatif Zaidi et al.

We investigate the generalization error of statistical learning models in a Federated Learning (FL) setting. Specifically, we study the evolution of the generalization error with the number of communication rounds $R$ between $K$ clients and a parameter server (PS), i.e., the effect on the generalization error of how often the clients' local models are aggregated at PS. In our setup, the more the clients communicate with PS the less data they use for local training in each round, such that the amount of training data per client is identical for distinct values of $R$. We establish PAC-Bayes and rate-distortion theoretic bounds on the generalization error that account explicitly for the effect of the number of rounds $R$, in addition to the number of participating devices $K$ and individual datasets size $n$. The bounds, which apply to a large class of loss functions and learning algorithms, appear to be the first of their kind for the FL setting. Furthermore, we apply our bounds to FL-type Support Vector Machines (FSVM); and derive (more) explicit bounds in this case. In particular, we show that the generalization bound of FSVM increases with $R$, suggesting that more frequent communication with PS diminishes the generalization power. This implies that the population risk decreases less fast with $R$ than does the empirical risk. Moreover, our bound suggests that the generalization error of FSVM decreases faster than that of centralized learning by a factor of $\mathcal{O}(\sqrt{\log(K)/K})$. Finally, we provide experimental results obtained using neural networks (ResNet-56) which show evidence that not only may our observations for FSVM hold more generally but also that the population risk may even start to increase beyond some value of $R$.

Milad Sefidgaran, Abdellatif Zaidi, Piotr Krasnowski

A major challenge in designing efficient statistical supervised learning algorithms is finding representations that perform well not only on available training samples but also on unseen data. While the study of representation learning has spurred much interest, most existing such approaches are heuristic; and very little is known about theoretical generalization guarantees. In this paper, we establish a compressibility framework that allows us to derive upper bounds on the generalization error of a representation learning algorithm in terms of the "Minimum Description Length" (MDL) of the labels or the latent variables (representations). Rather than the mutual information between the encoder's input and the representation, which is often believed to reflect the algorithm's generalization capability in the related literature but in fact, falls short of doing so, our new bounds involve the "multi-letter" relative entropy between the distribution of the representations (or labels) of the training and test sets and a fixed prior. In particular, these new bounds reflect the structure of the encoder and are not vacuous for deterministic algorithms. Our compressibility approach, which is information-theoretic in nature, builds upon that of Blum-Langford for PAC-MDL bounds and introduces two essential ingredients: block-coding and lossy-compression. The latter allows our approach to subsume the so-called geometrical compressibility as a special case. To the best knowledge of the authors, the established generalization bounds are the first of their kind for Information Bottleneck (IB) type encoders and representation learning. Finally, we partly exploit the theoretical results by introducing a new data-dependent prior. Numerical simulations illustrate the advantages of well-chosen such priors over classical priors used in IB.

Milad Sefidgaran, Abdellatif Zaidi, Piotr Krasnowski

We establish in-expectation and tail bounds on the generalization error of representation learning type algorithms. The bounds are in terms of the relative entropy between the distribution of the representations extracted from the training and "test'' datasets and a data-dependent symmetric prior, i.e., the Minimum Description Length (MDL) of the latent variables for the training and test datasets. Our bounds are shown to reflect the "structure" and "simplicity'' of the encoder and significantly improve upon the few existing ones for the studied model. We then use our in-expectation bound to devise a suitable data-dependent regularizer; and we investigate thoroughly the important question of the selection of the prior. We propose a systematic approach to simultaneously learning a data-dependent Gaussian mixture prior and using it as a regularizer. Interestingly, we show that a weighted attention mechanism emerges naturally in this procedure. Our experiments show that our approach outperforms the now popular Variational Information Bottleneck (VIB) method as well as the recent Category-Dependent VIB (CDVIB).

Milad Sefidgaran, Abdellatif Zaidi, Piotr Krasnowski

We study the problem of distributed multi-view representation learning. In this problem, $K$ agents observe each one distinct, possibly statistically correlated, view and independently extracts from it a suitable representation in a manner that a decoder that gets all $K$ representations estimates correctly the hidden label. In the absence of any explicit coordination between the agents, a central question is: what should each agent extract from its view that is necessary and sufficient for a correct estimation at the decoder? In this paper, we investigate this question from a generalization error perspective. First, we establish several generalization bounds in terms of the relative entropy between the distribution of the representations extracted from training and "test" datasets and a data-dependent symmetric prior, i.e., the Minimum Description Length (MDL) of the latent variables for all views and training and test datasets. Then, we use the obtained bounds to devise a regularizer; and investigate in depth the question of the selection of a suitable prior. In particular, we show and conduct experiments that illustrate that our data-dependent Gaussian mixture priors with judiciously chosen weights lead to good performance. For single-view settings (i.e., $K=1$), our experimental results are shown to outperform existing prior art Variational Information Bottleneck (VIB) and Category-Dependent VIB (CDVIB) approaches. Interestingly, we show that a weighted attention mechanism emerges naturally in this setting. Finally, for the multi-view setting, we show that the selection of the joint prior as a Gaussians product mixture induces a Gaussian mixture marginal prior for each marginal view and implicitly encourages the agents to extract and output redundant features, a finding which is somewhat counter-intuitive.

Masoud Kavian, Romain Chor, Milad Sefidgaran et al.

In this paper, we investigate the effect of data heterogeneity across clients on the performance of distributed learning systems, i.e., one-round Federated Learning, as measured by the associated generalization error. Specifically, $K$ clients have each $n$ training samples generated independently according to a possibly different data distribution, and their individually chosen models are aggregated by a central server. We study the effect of the discrepancy between the clients' data distributions on the generalization error of the aggregated model. First, we establish in-expectation and tail upper bounds on the generalization error in terms of the distributions. In part, the bounds extend the popular Conditional Mutual Information (CMI) bound, which was developed for the centralized learning setting, i.e., $K=1$, to the distributed learning setting with an arbitrary number of clients $K \geq 1$. Then, we connect with information-theoretic rate-distortion theory to derive possibly tighter \textit{lossy} versions of these bounds. Next, we apply our lossy bounds to study the effect of data heterogeneity across clients on the generalization error for the distributed classification problem in which each client uses Support Vector Machines (DSVM). In this case, we establish explicit generalization error bounds that depend explicitly on the data heterogeneity degree. It is shown that the bound gets smaller as the degree of data heterogeneity across clients increases, thereby suggesting that DSVM generalizes better when the dissimilarity between the clients' training samples is bigger. This finding, which goes beyond DSVM, is validated experimentally through several experiments.

Milad Sefidgaran, Kimia Nadjahi, Abdellatif Zaidi

In this paper, we leverage stochastic projection and lossy compression to establish new conditional mutual information (CMI) bounds on the generalization error of statistical learning algorithms. It is shown that these bounds are generally tighter than the existing ones. In particular, we prove that for certain problem instances for which existing MI and CMI bounds were recently shown in Attias et al. [2024] and Livni [2023] to become vacuous or fail to describe the right generalization behavior, our bounds yield suitable generalization guarantees of the order of $\mathcal{O}(1/\sqrt{n})$, where $n$ is the size of the training dataset. Furthermore, we use our bounds to investigate the problem of data "memorization" raised in those works, and which asserts that there are learning problem instances for which any learning algorithm that has good prediction there exist distributions under which the algorithm must "memorize" a big fraction of the training dataset. We show that for every learning algorithm, there exists an auxiliary algorithm that does not memorize and which yields comparable generalization error for any data distribution. In part, this shows that memorization is not necessary for good generalization.

Milad Sefidgaran, Abdellatif Zaidi, Piotr Krasnowski

A client device which has access to $n$ training data samples needs to obtain a statistical hypothesis or model $W$ and then to send it to a remote server. The client and the server devices share some common randomness sequence as well as a prior on the hypothesis space. In this problem a suitable hypothesis or model $W$ should meet two distinct design criteria simultaneously: (i) small (population) risk during the inference phase and (ii) small 'complexity' for it to be conveyed to the server with minimum communication cost. In this paper, we propose a joint training and source coding scheme with provable in-expectation guarantees, where the expectation is over the encoder's output message. Specifically, we show that by imposing a constraint on a suitable Kullback-Leibler divergence between the conditional distribution induced by a compressed learning model $\widehat{W}$ given $W$ and the prior, one guarantees simultaneously small average empirical risk (aka training loss), small average generalization error and small average communication cost. We also consider a one-shot scenario in which the guarantees on the empirical risk and generalization error are obtained for every encoder's output message.

Matei Moldoveanu, Abdellatif Zaidi

It is widely perceived that leveraging the success of modern machine learning techniques to mobile devices and wireless networks has the potential of enabling important new services. This, however, poses significant challenges, essentially due to that both data and processing power are highly distributed in a wireless network. In this paper, we develop a learning algorithm and an architecture that make use of multiple data streams and processing units, not only during the training phase but also during the inference phase. In particular, the analysis reveals how inference propagates and fuses across a network. We study the design criterion of our proposed method and its bandwidth requirements. Also, we discuss implementation aspects using neural networks in typical wireless radio access; and provide experiments that illustrate benefits over state-of-the-art techniques.

Septimia Sarbu, Abdellatif Zaidi

We consider the problem of learning parametric distributions from their quantized samples in a network. Specifically, $n$ agents or sensors observe independent samples of an unknown parametric distribution; and each of them uses $k$ bits to describe its observed sample to a central processor whose goal is to estimate the unknown distribution. First, we establish a generalization of the well-known van Trees inequality to general $L_p$-norms, with $p > 1$, in terms of Generalized Fisher information. Then, we develop minimax lower bounds on the estimation error for two losses: general $L_p$-norms and the related Wasserstein loss from optimal transport.

Matei Moldoveanu, Abdellatif Zaidi

In this paper, we consider a problem in which distributively extracted features are used for performing inference in wireless networks. We elaborate on our proposed architecture, which we herein refer to as "in-network learning", provide a suitable loss function and discuss its optimization using neural networks. We compare its performance with both Federated- and Split learning; and show that this architecture offers both better accuracy and bandwidth savings.

Mohammad Mahdi Mahvari, Mari Kobayashi, Abdellatif Zaidi

In the context of statistical learning, the Information Bottleneck method seeks a right balance between accuracy and generalization capability through a suitable tradeoff between compression complexity, measured by minimum description length, and distortion evaluated under logarithmic loss measure. In this paper, we study a variation of the problem, called scalable information bottleneck, in which the encoder outputs multiple descriptions of the observation with increasingly richer features. The model, which is of successive-refinement type with degraded side information streams at the decoders, is motivated by some application scenarios that require varying levels of accuracy depending on the allowed (or targeted) level of complexity. We establish an analytic characterization of the optimal relevance-complexity region for vector Gaussian sources. Then, we derive a variational inference type algorithm for general sources with unknown distribution; and show means of parametrizing it using neural networks. Finally, we provide experimental results on the MNIST dataset which illustrate that the proposed method generalizes better to unseen data during the training phase.

Mohammad Mahdi Mahvari, Mari Kobayashi, Abdellatif Zaidi

The Information Bottleneck method is a learning technique that seeks a right balance between accuracy and generalization capability through a suitable tradeoff between compression complexity, measured by minimum description length, and distortion evaluated under logarithmic loss measure. In this paper, we study a variation of the problem, called scalable information bottleneck, where the encoder outputs multiple descriptions of the observation with increasingly richer features. The problem at hand is motivated by some application scenarios that require varying levels of accuracy depending on the allowed level of generalization. First, we establish explicit (analytic) characterizations of the relevance-complexity region for memoryless Gaussian sources and memoryless binary sources. Then, we derive a Blahut-Arimoto type algorithm that allows us to compute (an approximation of) the region for general discrete sources. Finally, an application example in the pattern classification problem is provided along with numerical results.

Abdellatif Zaidi, Inaki Estella Aguerri, Shlomo Shamai

This tutorial paper focuses on the variants of the bottleneck problem taking an information theoretic perspective and discusses practical methods to solve it, as well as its connection to coding and learning aspects. The intimate connections of this setting to remote source-coding under logarithmic loss distortion measure, information combining, common reconstruction, the Wyner-Ahlswede-Korner problem, the efficiency of investment information, as well as, generalization, variational inference, representation learning, autoencoders, and others are highlighted. We discuss its extension to the distributed information bottleneck problem with emphasis on the Gaussian model and highlight the basic connections to the uplink Cloud Radio Access Networks (CRAN) with oblivious processing. For this model, the optimal trade-offs between relevance (i.e., information) and complexity (i.e., rates) in the discrete and vector Gaussian frameworks is determined. In the concluding outlook, some interesting problems are mentioned such as the characterization of the optimal inputs ("features") distributions under power limitations maximizing the "relevance" for the Gaussian information bottleneck, under "complexity" constraints.

Yigit Ugur, George Arvanitakis, Abdellatif Zaidi

In this paper, we develop an unsupervised generative clustering framework that combines the Variational Information Bottleneck and the Gaussian Mixture Model. Specifically, in our approach, we use the Variational Information Bottleneck method and model the latent space as a mixture of Gaussians. We derive a bound on the cost function of our model that generalizes the Evidence Lower Bound (ELBO) and provide a variational inference type algorithm that allows computing it. In the algorithm, the coders' mappings are parametrized using neural networks, and the bound is approximated by Monte Carlo sampling and optimized with stochastic gradient descent. Numerical results on real datasets are provided to support the efficiency of our method.

Inaki Estella Aguerri, Abdellatif Zaidi

The problem of distributed representation learning is one in which multiple sources of information $X_1,\ldots,X_K$ are processed separately so as to learn as much information as possible about some ground truth $Y$. We investigate this problem from information-theoretic grounds, through a generalization of Tishby's centralized Information Bottleneck (IB) method to the distributed setting. Specifically, $K$ encoders, $K \geq 2$, compress their observations $X_1,\ldots,X_K$ separately in a manner such that, collectively, the produced representations preserve as much information as possible about $Y$. We study both discrete memoryless (DM) and memoryless vector Gaussian data models. For the discrete model, we establish a single-letter characterization of the optimal tradeoff between complexity (or rate) and relevance (or information) for a class of memoryless sources (the observations $X_1,\ldots,X_K$ being conditionally independent given $Y$). For the vector Gaussian model, we provide an explicit characterization of the optimal complexity-relevance tradeoff. Furthermore, we develop a variational bound on the complexity-relevance tradeoff which generalizes the evidence lower bound (ELBO) to the distributed setting. We also provide two algorithms that allow to compute this bound: i) a Blahut-Arimoto type iterative algorithm which enables to compute optimal complexity-relevance encoding mappings by iterating over a set of self-consistent equations, and ii) a variational inference type algorithm in which the encoding mappings are parametrized by neural networks and the bound approximated by Markov sampling and optimized with stochastic gradient descent. Numerical results on synthetic and real datasets are provided to support the efficiency of the approaches and algorithms developed in this paper.