arXiv subjects

Anne-Marie Kermarrec

Publications and source records attributed to Anne-Marie Kermarrec.

At least 37 records · Page 2Linked to original sources

arXiv2025

Leveraging Approximate Caching for Faster Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) improves the reliability of large language model (LLM) answers by integrating external knowledge. However, RAG increases the end-to-end inference time since looking for relevant documents from large vector databases is computationally expensive. To address this, we introduce Proximity, an approximate key-value cache that optimizes the RAG workflow by leveraging similarities in user queries. Instead of treating each query independently, Proximity reuses previously retrieved documents when similar queries appear, substantially reducing the reliance on expensive vector database lookups. To efficiently scale, Proximity employs a locality-sensitive hashing (LSH) scheme that enables fast cache lookups while preserving retrieval accuracy. We evaluate Proximity using the MMLU and MedRAG question-answering benchmarks. Our experiments demonstrate that Proximity with our LSH scheme and a realistically-skewed MedRAG workload reduces database calls by 77.2% while maintaining database recall and test accuracy. We experiment with different similarity tolerances and cache capacities, and show that the time spent within the Proximity cache remains low and constant (4.8 microseconds) even as the cache grows substantially in size. Our results demonstrate that approximate caching is a practical and effective strategy for optimizing RAG-based systems.

cs.DB↗

arXiv2025

Efficient Pyramidal Analysis of Gigapixel Images on a Decentralized Modest Computer Cluster

Analyzing gigapixel images is recognized as computationally demanding. In this paper, we introduce PyramidAI, a technique for analyzing gigapixel images with reduced computational cost. The proposed approach adopts a gradual analysis of the image, beginning with lower resolutions and progressively concentrating on regions of interest for detailed examination at higher resolutions. We investigated two strategies for tuning the accuracy-computation performance trade-off when implementing the adaptive resolution selection, validated against the Camelyon16 dataset of biomedical images. Our results demonstrate that PyramidAI substantially decreases the amount of processed data required for analysis by up to 2.65x, while preserving the accuracy in identifying relevant sections on a single computer. To ensure democratization of gigapixel image analysis, we evaluated the potential to use mainstream computers to perform the computation by exploiting the parallelism potential of the approach. Using a simulator, we estimated the best data distribution and load balancing algorithm according to the number of workers. The selected algorithms were implemented and highlighted the same conclusions in a real-world setting. Analysis time is reduced from more than an hour to a few minutes using 12 modest workers, offering a practical solution for efficient large-scale image analysis.

cs.DC↗

arXiv2025

Fast In-Spectrum Graph Watermarks

We address the problem of watermarking graph objects, which consists in hiding information within them, to prove their origin. The two existing methods to watermark graphs use subgraph matching or graph isomorphism techniques, which are known to be intractable for large graphs. To reduce the operational complexity, we propose FFG, a new graph watermarking scheme adapted from an image watermarking scheme, since graphs and images can be represented as matrices. We analyze and compare FFG, whose novelty lies in embedding the watermark in the Fourier transform of the adjacency matrix of a graph. Our technique enjoys a much lower complexity than that of related works (i.e. in $\mathcal{O}\left(N^2 \log N\right)$), while performing better or at least as well as the two state-of-the-art methods.

cs.DS↗

arXiv2025

HarMoEny: Efficient Multi-GPU Inference of MoE Models

Mixture-of-Experts (MoE) models offer computational efficiency during inference by activating only a subset of specialized experts for a given input. This enables efficient model scaling on multi-GPU systems that use expert parallelism without compromising performance. However, load imbalance among experts and GPUs introduces waiting times, which can significantly increase inference latency. To address this challenge, we propose HarMoEny, a novel solution to address MoE load imbalance through two simple techniques: (i) dynamic token redistribution to underutilized GPUs and (ii) asynchronous prefetching of experts from the system to GPU memory. These techniques achieve a near-perfect load balance among experts and GPUs and mitigate delays caused by overloaded GPUs. We implement HarMoEny and compare its latency and throughput with four MoE baselines using real-world and synthetic datasets. Under heavy load imbalance, HarMoEny increases throughput by 37%-70% and reduces time-to-first-token by 34%-41%, compared to the next-best baseline. Moreover, our ablation study demonstrates that HarMoEny's scheduling policy reduces the GPU idling time by up to 84% compared to the baseline policies.

cs.DC↗

arXiv2025

Boosting Resource-Constrained Federated Learning Systems with Guessed Updates

Federated learning (FL) enables a set of client devices to collaboratively train a model without sharing raw data. This process, though, operates under the constrained computation and communication resources of edge devices. These constraints combined with systems heterogeneity force some participating clients to perform fewer local updates than expected by the server, thus slowing down convergence. Exhaustive tuning of hyperparameters in FL, furthermore, can be resource-intensive, without which the convergence is adversely affected. In this work, we propose GEL, the guess and learn algorithm. GEL enables constrained edge devices to perform additional learning through guessed updates on top of gradient-based steps. These guesses are gradientless, i.e., participating clients leverage them for free. Our generic guessing algorithm (i) can be flexibly combined with several state-of-the-art algorithms including FEDPROX, FEDNOVA, FEDYOGI or SCALEFL; and (ii) achieves significantly improved performance when the learning rates are not best tuned. We conduct extensive experiments and show that GEL can boost empirical convergence by up to 40% in resource constrained networks while relieving the need for exhaustive learning rate tuning.

cs.LG↗

arXiv2025

Robust ML Auditing using Prior Knowledge

Among the many technical challenges to enforcing AI regulations, one crucial yet underexplored problem is the risk of audit manipulation. This manipulation occurs when a platform deliberately alters its answers to a regulator to pass an audit without modifying its answers to other users. In this paper, we introduce a novel approach to manipulation-proof auditing by taking into account the auditor's prior knowledge of the task solved by the platform. We first demonstrate that regulators must not rely on public priors (e.g. a public dataset), as platforms could easily fool the auditor in such cases. We then formally establish the conditions under which an auditor can prevent audit manipulations using prior knowledge about the ground truth. Finally, our experiments with two standard datasets illustrate the maximum level of unfairness a platform can hide before being detected as malicious. Our formalization and generalization of manipulation-proof auditing with a prior opens up new research directions for more robust fairness audits.

cs.LG↗

arXiv2025

Low-Cost Privacy-Preserving Decentralized Learning

Decentralized learning (DL) is an emerging paradigm of collaborative machine learning that enables nodes in a network to train models collectively without sharing their raw data or relying on a central server. This paper introduces Zip-DL, a privacy-aware DL algorithm that leverages correlated noise to achieve robust privacy against local adversaries while ensuring efficient convergence at low communication costs. By progressively neutralizing the noise added during distributed averaging, Zip-DL combines strong privacy guarantees with high model accuracy. Its design requires only one communication round per gradient descent iteration, significantly reducing communication overhead compared to competitors. We establish theoretical bounds on both convergence speed and privacy guarantees. Moreover, extensive experiments demonstrating Zip-DL's practical applicability make it outperform state-of-the-art methods in the accuracy vs. vulnerability trade-off. Specifically, Zip-DL (i) reduces membership-inference attack success rates by up to 35% compared to baseline DL, (ii) decreases attack efficacy by up to 13% compared to competitors offering similar utility, and (iii) achieves up to 59% higher accuracy to completely nullify a basic attack scenario, compared to a state-of-the-art privacy-preserving approach under the same threat model. These results position Zip-DL as a practical and efficient solution for privacy-preserving decentralized learning in real-world applications.

cs.LG↗

arXiv2025

Harnessing Increased Client Participation with Cohort-Parallel Federated Learning

Federated learning (FL) is a machine learning approach where nodes collaboratively train a global model. As more nodes participate in a round of FL, the effectiveness of individual model updates by nodes also diminishes. In this study, we increase the effectiveness of client updates by dividing the network into smaller partitions, or cohorts. We introduce Cohort-Parallel Federated Learning (CPFL): a novel learning approach where each cohort independently trains a global model using FL, until convergence, and the produced models by each cohort are then unified using knowledge distillation. The insight behind CPFL is that smaller, isolated networks converge quicker than in a one-network setting where all nodes participate. Through exhaustive experiments involving realistic traces and non-IID data distributions on the CIFAR-10 and FEMNIST image classification tasks, we investigate the balance between the number of cohorts, model accuracy, training time, and compute resources. Compared to traditional FL, CPFL with four cohorts, non-IID data distribution, and CIFAR-10 yields a 1.9x reduction in train time and a 1.3x reduction in resource usage, with a minimal drop in test accuracy.

cs.LG↗

arXiv2025

Practical Federated Learning without a Server

Federated Learning (FL) enables end-user devices to collaboratively train ML models without sharing raw data, thereby preserving data privacy. In FL, a central parameter server coordinates the learning process by iteratively aggregating the trained models received from clients. Yet, deploying a central server is not always feasible due to hardware unavailability, infrastructure constraints, or operational costs. We present Plexus, a fully decentralized FL system for large networks that operates without the drawbacks originating from having a central server. Plexus distributes the responsibilities of model aggregation and sampling among participating nodes while avoiding network-wide coordination. We evaluate Plexus using realistic traces for compute speed, pairwise latency and network capacity. Our experiments on three common learning tasks and with up to 1000 nodes empirically show that Plexus reduces time-to-accuracy by 1.4-1.6x, communication volume by 15.8-292x and training resources needed for convergence by 30.5-77.9x compared to conventional decentralized learning algorithms.

cs.DC↗

arXiv2025

Boosting Asynchronous Decentralized Learning with Model Fragmentation

Decentralized learning (DL) is an emerging technique that allows nodes on the web to collaboratively train machine learning models without sharing raw data. Dealing with stragglers, i.e., nodes with slower compute or communication than others, is a key challenge in DL. We present DivShare, a novel asynchronous DL algorithm that achieves fast model convergence in the presence of communication stragglers. DivShare achieves this by having nodes fragment their models into parameter subsets and send, in parallel to computation, each subset to a random sample of other nodes instead of sequentially exchanging full models. The transfer of smaller fragments allows more efficient usage of the collective bandwidth and enables nodes with slow network links to quickly contribute with at least some of their model parameters. By theoretically proving the convergence of DivShare, we provide, to the best of our knowledge, the first formal proof of convergence for a DL algorithm that accounts for the effects of asynchronous communication with delays. We experimentally evaluate DivShare against two state-of-the-art DL baselines, AD-PSGD and Swift, and with two standard datasets, CIFAR-10 and MovieLens. We find that DivShare with communication stragglers lowers time-to-accuracy by up to 3.9x compared to AD-PSGD on the CIFAR-10 dataset. Compared to baselines, DivShare also achieves up to 19.4% better accuracy and 9.5% lower test loss on the CIFAR-10 and MovieLens datasets, respectively.

cs.DC↗

arXiv2025

Fair Decentralized Learning

Decentralized learning (DL) is an emerging approach that enables nodes to collaboratively train a machine learning model without sharing raw data. In many application domains, such as healthcare, this approach faces challenges due to the high level of heterogeneity in the training data's feature space. Such feature heterogeneity lowers model utility and negatively impacts fairness, particularly for nodes with under-represented training data. In this paper, we introduce \textsc{Facade}, a clustering-based DL algorithm specifically designed for fair model training when the training data exhibits several distinct features. The challenge of \textsc{Facade} is to assign nodes to clusters, one for each feature, based on the similarity in the features of their local data, without requiring individual nodes to know apriori which cluster they belong to. \textsc{Facade} (1) dynamically assigns nodes to their appropriate clusters over time, and (2) enables nodes to collaboratively train a specialized model for each cluster in a fully decentralized manner. We theoretically prove the convergence of \textsc{Facade}, implement our algorithm, and compare it against three state-of-the-art baselines. Our experimental results on three datasets demonstrate the superiority of our approach in terms of model accuracy and fairness compared to all three competitors. Compared to the best-performing baseline, \textsc{Facade} on the CIFAR-10 dataset also reduces communication costs by 32.3\% to reach a target accuracy when cluster sizes are imbalanced.

cs.LG↗

arXiv2024

Revisiting Ensembling in One-Shot Federated Learning

Federated learning (FL) is an appealing approach to training machine learning models without sharing raw data. However, standard FL algorithms are iterative and thus induce a significant communication cost. One-shot federated learning (OFL) trades the iterative exchange of models between clients and the server with a single round of communication, thereby saving substantially on communication costs. Not surprisingly, OFL exhibits a performance gap in terms of accuracy with respect to FL, especially under high data heterogeneity. We introduce FENS, a novel federated ensembling scheme that approaches the accuracy of FL with the communication efficiency of OFL. Learning in FENS proceeds in two phases: first, clients train models locally and send them to the server, similar to OFL; second, clients collaboratively train a lightweight prediction aggregator model using FL. We showcase the effectiveness of FENS through exhaustive experiments spanning several datasets and heterogeneity levels. In the particular case of heterogeneously distributed CIFAR-10 dataset, FENS achieves up to a 26.9% higher accuracy over state-of-the-art (SOTA) OFL, being only 3.1% lower than FL. At the same time, FENS incurs at most 4.3x more communication than OFL, whereas FL is at least 10.9x more communication-intensive than FENS.

cs.LG↗

arXiv2024

PeerSwap: A Peer-Sampler with Randomness Guarantees

The ability of a peer-to-peer (P2P) system to effectively host decentralized applications often relies on the availability of a peer-sampling service, which provides each participant with a random sample of other peers. Despite the practical effectiveness of existing peer samplers, their ability to produce random samples within a reasonable time frame remains poorly understood from a theoretical standpoint. This paper contributes to bridging this gap by introducing PeerSwap, a peer-sampling protocol with provable randomness guarantees. We establish execution time bounds for PeerSwap, demonstrating its ability to scale effectively with the network size. We prove that PeerSwap maintains the fixed structure of the communication graph while allowing sequential peer position swaps within this graph. We do so by showing that PeerSwap is a specific instance of an interchange process, a renowned model for particle movement analysis. Leveraging this mapping, we derive execution time bounds, expressed as a function of the network size N. Depending on the network structure, this time can be as low as a polylogarithmic function of N, highlighting the efficiency of PeerSwap. We implement PeerSwap and conduct numerical evaluations using regular graphs with varying connectivity and containing up to 32768 (2^15) peers. Our evaluation demonstrates that PeerSwap quickly provides peers with uniform random samples of other peers.

cs.DC↗

arXiv2024

Noiseless Privacy-Preserving Decentralized Learning

Decentralized learning (DL) enables collaborative learning without a server and without training data leaving the users' devices. However, the models shared in DL can still be used to infer training data. Conventional defenses such as differential privacy and secure aggregation fall short in effectively safeguarding user privacy in DL, either sacrificing model utility or efficiency. We introduce Shatter, a novel DL approach in which nodes create virtual nodes (VNs) to disseminate chunks of their full model on their behalf. This enhances privacy by (i) preventing attackers from collecting full models from other nodes, and (ii) hiding the identity of the original node that produced a given model chunk. We theoretically prove the convergence of Shatter and provide a formal analysis demonstrating how Shatter reduces the efficacy of attacks compared to when exchanging full models between nodes. We evaluate the convergence and attack resilience of Shatter with existing DL algorithms, with heterogeneous datasets, and against three standard privacy attacks. Our evaluation shows that Shatter not only renders these privacy attacks infeasible when each node operates 16 VNs but also exhibits a positive impact on model utility compared to standard DL. In summary, Shatter enhances the privacy of DL while maintaining the utility and efficiency of the model.

cs.DC↗

arXiv2024

Fairness Auditing with Multi-Agent Collaboration

Existing work in fairness auditing assumes that each audit is performed independently. In this paper, we consider multiple agents working together, each auditing the same platform for different tasks. Agents have two levers: their collaboration strategy, with or without coordination beforehand, and their strategy for sampling appropriate data points. We theoretically compare the interplay of these levers. Our main findings are that (i) collaboration is generally beneficial for accurate audits, (ii) basic sampling methods often prove to be effective, and (iii) counter-intuitively, extensive coordination on queries often deteriorates audits accuracy as the number of agents increases. Experiments on three large datasets confirm our theoretical results. Our findings motivate collaboration during fairness audits of platforms that use ML models for decision-making.

cs.LG↗

arXiv2024

Energy-Aware Decentralized Learning with Intermittent Model Training

Decentralized learning (DL) offers a powerful framework where nodes collaboratively train models without sharing raw data and without the coordination of a central server. In the iterative rounds of DL, models are trained locally, shared with neighbors in the topology, and aggregated with other models received from neighbors. Sharing and merging models contribute to convergence towards a consensus model that generalizes better across the collective data captured at training time. In addition, the energy consumption while sharing and merging model parameters is negligible compared to the energy spent during the training phase. Leveraging this fact, we present SkipTrain, a novel DL algorithm, which minimizes energy consumption in decentralized learning by strategically skipping some training rounds and substituting them with synchronization rounds. These training-silent periods, besides saving energy, also allow models to better mix and finally produce models with superior accuracy than typical DL algorithms that train at every round. Our empirical evaluations with 256 nodes demonstrate that SkipTrain reduces energy consumption by 50% and increases model accuracy by up to 12% compared to D-PSGD, the conventional DL algorithm.

cs.LG↗

arXiv2024

Decentralized Learning Made Practical with Client Sampling

Decentralized learning (DL) leverages edge devices for collaborative model training while avoiding coordination by a central server. Due to privacy concerns, DL has become an attractive alternative to centralized learning schemes since training data never leaves the device. In a round of DL, all nodes participate in model training and exchange their model with some other nodes. Performing DL in large-scale heterogeneous networks results in high communication costs and prolonged round durations due to slow nodes, effectively inflating the total training time. Furthermore, current DL algorithms also assume all nodes are available for training and aggregation at all times, diminishing the practicality of DL. This paper presents Plexus, an efficient, scalable, and practical DL system. Plexus (1) avoids network-wide participation by introducing a decentralized peer sampler that selects small subsets of available nodes that train the model each round and, (2) aggregates the trained models produced by nodes every round. Plexus is designed to handle joining and leaving nodes (churn). We extensively evaluate Plexus by incorporating realistic traces for compute speed, pairwise latency, network capacity, and availability of edge devices in our experiments. Our experiments on four common learning tasks empirically show that Plexus reduces time-to-accuracy by 1.2-8.3x, communication volume by 2.4-15.3x and training resources needed for convergence by 6.4-370x compared to baseline DL algorithms.

cs.DC↗

arXiv2023

Get More for Less in Decentralized Learning Systems

Decentralized learning (DL) systems have been gaining popularity because they avoid raw data sharing by communicating only model parameters, hence preserving data confidentiality. However, the large size of deep neural networks poses a significant challenge for decentralized training, since each node needs to exchange gigabytes of data, overloading the network. In this paper, we address this challenge with JWINS, a communication-efficient and fully decentralized learning system that shares only a subset of parameters through sparsification. JWINS uses wavelet transform to limit the information loss due to sparsification and a randomized communication cut-off that reduces communication usage without damaging the performance of trained models. We demonstrate empirically with 96 DL nodes on non-IID datasets that JWINS can achieve similar accuracies to full-sharing DL while sending up to 64% fewer bytes. Additionally, on low communication budgets, JWINS outperforms the state-of-the-art communication-efficient DL algorithm CHOCO-SGD by up to 4x in terms of network savings and time.

cs.DC↗