Interpretable Basis

Interpretable basis research aims to create models that not only produce accurate predictions but also offer transparent explanations for their decisions. Current efforts focus on developing methods to learn and utilize these bases within various architectures, including attention-based models and dynamic ensembles, often employing techniques like contrastive learning and Kashin quantization for improved efficiency and interpretability. This work is significant because it addresses the "black box" nature of many machine learning models, enhancing trust and facilitating better understanding of complex systems in diverse fields like time series forecasting and image classification. Improved interpretability ultimately leads to more reliable and trustworthy AI systems.

Papers

April 15, 2024

Quantization of Large Language Models with an Overdetermined Basis
Daniil Merkulov, Daria Cherniuk, Alexander Rudikov, Ivan Oseledets, Ekaterina Muravleva, Aleksandr Mikhalev, Boris Kashin
Quantization Operator Data Compression Compression Algorithm Interpretable Basis

January 3, 2024

Towards Modeling Uncertainties of Self-explaining Neural Networks via Conformal Prediction
Wei Qian, Chenxu Zhao, Yangyi Li, Fenglong Ma, Chao Zhang, Mengdi Huai
Conformal Prediction Uncertainty Modeling Free Uncertainty Built in Interpretability Self Explaining Neural Network Interpretable Basis

October 31, 2023

BasisFormer: Attention-based Time Series Forecasting with Learnable and Interpretable Basis
Zelin Ni, Hang Yu, Shizhan Liu, Jianguo Li, Weiyao Lin
Time Series Forecasting Multivariate Time Series Forecasting Interpretable Basis

March 19, 2023

Unsupervised Interpretable Basis Extraction for Concept-Based Visual Explanations
Alexandros Doumanoglou, Stylianos Asteriadis, Dimitrios Zarpalas
Latent Direction Mid Level Representation Interpretable Basis

January 14, 2022

IDEA: Interpretable Dynamic Ensemble Architecture for Time Series Prediction
Mengyue Zha, Kani Chen, Tong Zhang
Time Series Prediction Univariate Time Series Forecasting Competitive Learning Interpretable Ensemble Explainable Ensemble Interpretable Basis

Interpretable Basis

Papers

Quantization of Large Language Models with an Overdetermined Basis

Towards Modeling Uncertainties of Self-explaining Neural Networks via Conformal Prediction

BasisFormer: Attention-based Time Series Forecasting with Learnable and Interpretable Basis

Unsupervised Interpretable Basis Extraction for Concept-Based Visual Explanations

IDEA: Interpretable Dynamic Ensemble Architecture for Time Series Prediction