Leveraging Periodicity for Tabular Deep Learning

Electronics, Journal Year: 2025, Volume and Issue: 14(6), P. 1165 - 1165

Published: March 16, 2025

Deep learning has achieved remarkable success in various domains; however, its application to tabular data remains challenging due the complex nature of feature interactions and patterns. This paper introduces novel neural network architectures that leverage intrinsic periodicity enhance prediction accuracy for regression classification tasks. We propose FourierNet, which employs a Fourier-based encoder capture periodic patterns, ChebyshevNet, utilizing Chebyshev-based model non-periodic Furthermore, we combine these approaches two architectures: Periodic-Non-Periodic Network (PNPNet) AutoPNPNet. PNPNet detects features priori, feeding them into separate branches, while AutoPNPNet automatically selects through learned mechanism. The experimental results on benchmark 53 datasets demonstrate our methods outperform current state-of-the-art deep technique 34 show interesting properties explainability.

Language: Английский

A Literature Review of Explainable Tabular Data

Published: Aug. 8, 2024

Explainable Artificial Intelligence (XAI) plays a vital role in increasing transparency and trust machine learning models, particularly when applied to tabular data which is used domains such as finance, healthcare, marketing. This paper presents an extensive survey of XAI techniques with aims analyze recent research developments since 2021. The classes describes several pertinent data, it identifies challenges specific this domain, explores potential applications emerging trends. Future directions are outlined, concentrating on the need for clear definitions terminology used, security, user-centric explanations, enhanced interaction, robust evaluation metrics, advancements adversarial example-based analysis. aim contribute evolving field XAI, provide insights effective, trustworthy, transparent decision-making using data.

Language: Английский