analiza aiqbench
FreqX introduces a novel interpretability method for personalized federated learning that leverages signal processing and information theory to deliver both attribution and concept insights. By operating at least 10 times faster than existing baselines while maintaining privacy, this approach addresses critical gaps in handling non-IID data and heterogeneous devices. The technique offers a significant efficiency gain for deploying explainable AI in resource-constrained, privacy-sensitive environments.
Comprehensive and Reliable Feature Attribution for Diverse Modalities and Models via Frequency-Domain Insights
arXiv:2411.18343v3 Announce Type: replace-cross Abstract: Personalized Federal learning(PFL) allows clients to cooperatively train a personalized model without disclosing their private dataset. However, PFL suffers from Non-IID, heterogeneous devices, lack of fairness, and unclear contribution which urgently need the interpretability of deep learning model to overcome these challenges. These challenges proposed new demands for interpretability. Low cost, privacy, and detailed information. There is no current interpretability method satisfying them. In this paper, we propose a novel interpretability method \emph{FreqX} by introducing Signal Processing and Information Theory. Our experiments show that the explanation results of FreqX contain both attribution information and concept information. FreqX runs at least 10 times faster than the baselines which contain concept information.
Abstract:Personalized Federal learning(PFL) allows clients to cooperatively train a personalized model without disclosing their private dataset. However, PFL suffers from Non-IID, heterogeneous devices, lack of fairness, and unclear contribution which urgently need the interpretability of deep learning model to overcome these challenges. These challenges proposed new demands for interpretability. Low cost, privacy, and detailed information. There is no current interpretability method satisfying them. In this paper, we propose a novel interpretability method \emph{FreqX} by introducing Signal Processing and Information Theory. Our experiments show that the explanation results of FreqX contain both attribution information and concept information. FreqX runs at least 10 times faster than the baselines which contain concept information.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2411.18343 [cs.LG]
(or arXiv:2411.18343v3 [cs.LG] for this version)
Text integral sub licența deschisă a publicației · 124 cuvinte