arXiv:2503.19209v3 Announce Type: replace 
Abstract: In this paper, we propose BR-MTRL, a Byzantine-resilient multi-task representation learning framework that handles faulty or malicious agents. Our approach leverages representation learning through a shared neural network model, where all clients share fixed layers, except for a client-specific final layer. This structure captures shared features among clients while enabling individual adaptation, making it a promising approach for leveraging client data and computational power in heterogeneous federated settings to learn personalized models. To learn the model, we employ an alternating gradient descent strategy: each client optimizes its local model, updates its final layer, and sends estimates of the shared representation to a central server for aggregation. To defend against Byzantine agents, we employ two robust aggregation methods for client-server communication, Geometric Median and Krum. Our method enables personalized learning while maintaining resilience in distributed settings. We implemented the proposed algorithm in a federated testbed built using Amazon Web Services (AWS) platform and compared its performance with various benchmark algorithms and their variations. Through experiments using real-world datasets, including CIFAR-10 and FEMNIST, we demonstrated the effectiveness and robustness of our approach and its transferability to new unseen clients with limited data, even in the presence of Byzantine adversaries.

تقديم إطار التعلم متعدد المهام المقاوم للبيزنطيين (BR-MTRL) يمثل تقدمًا كبيرًا في مجال التعلم الآلي، خاصة في البيئات التي تحتوي على وكلاء غير موثوقين. من خلال السماح للعملاء بمشاركة شبكة عصبية مشتركة مع الحفاظ على التكيفات الفردية، لا يعزز هذا النهج المبتكر فقط متانة أنظمة التعلم، بل يفتح أيضًا آفاقًا جديدة للمهام التعاونية. هذا أمر حاسم لأنه يعالج التحديات التي تطرحها الوكلاء المعطوبين أو الخبيثين، مما يضمن نتائج أكثر موثوقية وفعالية في سيناريوهات التعلم متعدد المهام.

La introducción del marco de Aprendizaje de Representación Multi-Tarea Resiliente a Bizancio (BR-MTRL) representa un avance significativo en el aprendizaje automático, especialmente en entornos con agentes poco fiables. Al permitir que los clientes compartan una red neuronal común mientras mantienen adaptaciones individuales, este enfoque innovador no solo mejora la robustez de los sistemas de aprendizaje, sino que también abre nuevas posibilidades para tareas colaborativas. Esto es crucial, ya que aborda los desafíos planteados por agentes defectuosos o maliciosos, asegurando resultados más confiables y efectivos en escenarios de aprendizaje multi-tarea.

L'introduction du cadre d'apprentissage de représentation multi-tâches résilient aux Byzantins (BR-MTRL) marque une avancée significative dans le domaine de l'apprentissage automatique, en particulier dans des environnements avec des agents peu fiables. En permettant aux clients de partager un réseau de neurones commun tout en maintenant des adaptations individuelles, cette approche innovante améliore non seulement la robustesse des systèmes d'apprentissage, mais ouvre également de nouvelles possibilités pour des tâches collaboratives. Cela est crucial car cela répond aux défis posés par des agents défectueux ou malveillants, garantissant des résultats plus fiables et efficaces dans les scénarios d'apprentissage multi-tâches.

The introduction of the Byzantine Resilient Multi-Task Representation Learning (BR-MTRL) framework marks a significant advancement in machine learning, particularly in environments with unreliable agents. By allowing clients to share a common neural network while maintaining individual adaptations, this innovative approach not only enhances the robustness of learning systems but also opens up new possibilities for collaborative tasks. This is crucial as it addresses the challenges posed by faulty or malicious agents, ensuring more reliable and effective outcomes in multi-task learning scenarios.

Byzantine Resilient Federated Multi-Task Representation Learning

arXiv:2601.08493v1 Announce Type: new 
Abstract: Few-shot class-incremental learning (FSCIL) aims to continually adapt a model on a limited number of new-class examples, facing two well-known challenges: catastrophic forgetting and overfitting to new classes. Existing methods tend to freeze more parts of network components and finetune others with an extra memory during incremental sessions. These methods emphasize preserving prior knowledge to ensure proficiency in recognizing old classes, thereby mitigating catastrophic forgetting. Meanwhile, constraining fewer parameters can help in overcoming overfitting with the assistance of prior knowledge. Following previous methods, we retain more prior knowledge and propose a prior knowledge-infused neural network (PKI) to facilitate FSCIL. PKI consists of a backbone, an ensemble of projectors, a classifier, and an extra memory. In each incremental session, we build a new projector and add it to the ensemble. Subsequently, we finetune the new projector and the classifier jointly with other frozen network components, ensuring the rich prior knowledge is utilized effectively. By cascading projectors, PKI integrates prior knowledge accumulated from previous sessions and learns new knowledge flexibly, which helps to recognize old classes and efficiently learn new classes. Further, to reduce the resource consumption associated with keeping many projectors, we design two variants of the prior knowledge-infused neural network (PKIV-1 and PKIV-2) to trade off a balance between resource consumption and performance by reducing the number of projectors. Extensive experiments on three popular benchmarks demonstrate that our approach outperforms state-of-the-art methods.

تم تقديم نهج جديد للتعلم التزايدي لفئات قليلة من الأمثلة (FSCIL) من خلال الشبكة العصبية المدعومة بالمعرفة السابقة (PKI)، والتي تهدف إلى تحسين تكيف النموذج مع عدد محدود من أمثلة الفئات الجديدة، مع مواجهة النسيان الكارثي والإفراط في التكيف. تستخدم PKI مجموعة من أجهزة العرض وذاكرة إضافية للاحتفاظ بالمعرفة السابقة بشكل فعال خلال جلسات التعلم التزايدي.

Se ha introducido un nuevo enfoque para el Aprendizaje Incremental de Clases con Pocos Ejemplos (FSCIL) a través de la Red Neuronal Infundida con Conocimiento Previo (PKI), que busca mejorar la adaptabilidad del modelo con un número limitado de ejemplos de nuevas clases, enfrentando el olvido catastrófico y el sobreajuste. PKI emplea un conjunto de proyectores y una memoria adicional para retener eficazmente el conocimiento previo durante las sesiones de aprendizaje incremental.

Une nouvelle approche de l'apprentissage incrémental de classe avec peu d'exemples (FSCIL) a été introduite grâce au réseau de neurones infusé de connaissances antérieures (PKI), qui vise à améliorer l'adaptabilité des modèles avec un nombre limité d'exemples de nouvelles classes tout en abordant l'oubli catastrophique et le surapprentissage. Le PKI utilise un ensemble de projecteurs et une mémoire supplémentaire pour conserver efficacement les connaissances antérieures lors des sessions d'apprentissage incrémental.

A new approach to Few-Shot Class-Incremental Learning (FSCIL) has been introduced through the Prior Knowledge-Infused Neural Network (PKI), which aims to enhance model adaptability with limited new-class examples while addressing catastrophic forgetting and overfitting. PKI employs an ensemble of projectors and an extra memory to retain prior knowledge effectively during incremental learning sessions.

PKI: Prior Knowledge-Infused Neural Network for Few-Shot Class-Incremental Learning

arXiv:2601.08760v1 Announce Type: new 
Abstract: We study a decentralized collaborative requesting problem that aims to optimize the information freshness of time-sensitive clients in edge networks consisting of multiple clients, access nodes (ANs), and servers. Clients request content through ANs acting as gateways, without observing AN states or the actions of other clients. We define the reward as the age of information reduction resulting from a client's selection of an AN, and formulate the problem as a non-stationary multi-armed bandit. In this decentralized and partially observable setting, the resulting reward process is history-dependent and coupled across clients, and exhibits both abrupt and gradual changes in expected rewards, rendering classical bandit-based approaches ineffective. To address these challenges, we propose the AGING BANDIT WITH ADAPTIVE RESET algorithm, which combines adaptive windowing with periodic monitoring to track evolving reward distributions. We establish theoretical performance guarantees showing that the proposed algorithm achieves near-optimal performance, and we validate the theoretical results through simulations.

تدرس دراسة جديدة نُشرت على arXiv مشكلة طلب تعاوني لامركزي تهدف إلى تحسين حداثة المعلومات للعملاء الحساسين للوقت في الشبكات الحافة. تقدم البحث خوارزمية AGING BANDIT WITH ADAPTIVE RESET، التي تتناول تحديات المكافآت المعتمدة على التاريخ في إطار متعدد الأذرع غير الثابت.

Un nuevo estudio publicado en arXiv investiga un problema de solicitud colaborativa descentralizada que busca optimizar la frescura de la información para clientes sensibles al tiempo en redes de borde. La investigación presenta el algoritmo AGING BANDIT WITH ADAPTIVE RESET, que aborda los desafíos de las recompensas dependientes de la historia en un marco de bandido multi-brazo no estacionario.

Une nouvelle étude publiée sur arXiv examine un problème de demande collaborative décentralisée visant à optimiser la fraîcheur de l'information pour des clients sensibles au temps dans des réseaux de périphérie. La recherche introduit l'algorithme AGING BANDIT WITH ADAPTIVE RESET, qui aborde les défis des récompenses dépendantes de l'historique dans un cadre de bandit multi-bras non stationnaire.

A new study published on arXiv investigates a decentralized collaborative requesting problem aimed at optimizing information freshness for time-sensitive clients in edge networks. The research introduces the AGING BANDIT WITH ADAPTIVE RESET algorithm, which addresses the challenges of history-dependent rewards in a non-stationary multi-armed bandit framework.

Byzantine Resilient Federated Multi-Task Representation Learning

Was this article worth reading? Share it

Chattermate

Humanize AI

LucidQuery AI

Https

AskTuring

Tribela

Ready to build your own newsroom?