FedAdaVR: Adaptive Variance Reduction for Robust Federated Learning under Limited Client Participation

⋅

Federated learning (FL) encounters substantial challenges due to heterogeneity, leading to gradient noise, client drift, and partial client participation errors, the last of which is the most pervasive but remains insufficiently addressed in current literature. In this paper, we propose FedAdaVR, a novel FL algorithm aimed at solving heterogeneity issues caused by sporadic client participation by incorporating an adaptive optimiser with a variance reduction technique. This method takes advantage of the most recent stored updates from clients, even […]

Ver mais

Liked Liked

Frozen Policy Iteration: Computationally Efficient RL under Linear $Q^{pi}$ Realizability for Deterministic Dynamics

⋅

arXiv:2603.00716v1 Announce Type: cross Abstract: We study computationally and statistically efficient reinforcement learning under the linear $Q^{pi}$ realizability assumption, where any policy’s $Q$-function is linear in a given state-action feature representation. Prior methods in this setting are either computationally intractable, or require (local) access to a simulator. In this paper, we propose a computationally efficient online RL algorithm, named Frozen Policy Iteration, under the linear $Q^{pi}$ realizability setting that works for Markov Decision Processes (MDPs) with stochastic initial […]

Ver mais

Liked Liked
When AI Finally Learned That “Dog” and 🐕 Are the Same Thing, aka CLIP

When AI Finally Learned That “Dog” and 🐕 Are the Same Thing, aka CLIP

⋅

Author(s): DrSwarnenduAI Originally published on Towards AI. How CLIP used 400 million internet image-caption pairs to solve the 60-year problem of connecting vision and language by making them occupy the same 512-dimensional manifold. Welcome back. I believe in coordinates and manifolds. If this 15-minute mathematical deep dive helps you, please leave a comment. I write these for the community, and your insights are what keep this series going. Image CaptionThe article delves into CLIP, a model that revolutionizes […]

Ver mais

Liked Liked
[R] Controlled LLM Training on Spectral Sphere

[R] Controlled LLM Training on Spectral Sphere

⋅

TL;DR: The paper introduces Spectral Sphere Optimizer, which takes steepest descent under spectral norm (Muon) and forces the weights & updates onto a spectral sphere. Paper: https://www.arxiv.org/pdf/2601.08393 Repo: https://github.com/Unakar/Spectral-Sphere-Optimizer Abstract: Scaling large models requires optimization strategies that ensure rapid convergence grounded in stability. Maximal Update Parametrization ( muP) provides a theoretical safeguard for width-invariant theta(1) activation control, whereas emerging optimizers like Muon are only “half-aligned” with these constraints: they control updates but allow weights to drift. To address […]

Ver mais

Liked Liked

Lost in Execution: On the Multilingual Robustness of Tool Calling in Large Language Models

⋅

arXiv:2601.05366v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly deployed as agents that invoke external tools through structured function calls. While recent work reports strong tool-calling performance under standard English-centric evaluations, the robustness of tool calling under multilingual user interactions remains underexplored. In this work, we introduce MLCL, a diagnostic benchmark, and conduct a systematic evaluation of multilingual tool calling across Chinese, Hindi, and the low-resource language Igbo. Through fine-grained error analysis, we show that many […]

Ver mais

Liked Liked

From Noise to Order: Learning to Rank via Denoising Diffusion

⋅

In information retrieval (IR), learning-to-rank (LTR) methods have traditionally limited themselves to discriminative machine learning approaches that model the probability of the document being relevant to the query given some feature representation of the query-document pair. In this work, we propose an alternative denoising diffusion-based deep generative approach to LTR that instead models the full joint distribution over feature vectors and relevance labels. While in the discriminative setting, an over-parameterized ranking model may find different ways to fit […]

Ver mais

Liked Liked

Pseudo Contrastive Learning for Diagram Comprehension in Multimodal Models

⋅

arXiv:2602.23589v1 Announce Type: new Abstract: Recent multimodal models such as Contrastive Language-Image Pre-training (CLIP) have shown remarkable ability to align visual and linguistic representations. However, domains where small visual differences carry large semantic significance, such as diagram understanding, remain challenging due to the models’ limited sensitivity to fine-grained structural variations. We propose a new training paradigm designed to enhance diagram comprehension in vision-language models. Our approach introduces pseudo contrastive samples generated by a diagram renderer that creates synthetic […]

Ver mais

Liked Liked

Quantum-Enhanced Adaptive Graph Convolutional Networks for Sentiment Representation Learning

⋅

Sentiment classification struggles with complex semantic relationships using static text graphs. We introduce the Quantum-Enhanced Adaptive Graph Convolutional Network (QAGCN), a hybrid quantum-classical architecture for robust sentiment representation. QAGCN’s core is a Quantum-Enhanced Graph Construction Module employing a Parameterized Quantum Circuit (PQC) to dynamically learn emotional association strengths between word pairs. This generates a task-adaptive adjacency matrix, which then feeds into classical GNN layers. Evaluations on benchmark datasets (Yelp, IMDB, Amazon, MC, RP) demonstrate QAGCN’s superior or competitive […]

Ver mais

Liked Liked

Color quantization using K-means clustering in ML.NET

⋅

When I was looking for K-means use cases, I found out about Color quantization, a very interesting . I implemented it in Python and was wondering whether it would be as easy to implement in ML.NET. All the code is available in this GitHub repository. What is color quantization Color quantization is the usage of quantization, a lossy compression technique, in color spaces in order to reduce the number of unique colors in an image.

Ver mais

Liked Liked

Load More

↑

Digitado © 2025