LatentExplainer: Explaining Latent Representations in Deep Generative Models with Multi-modal Foundation Models

深度生成模型，如VAEs和扩散模型，通过利用潜在变量学习数据分布并生成高质量样本，推进了各种生成任务。尽管可解释的AI领域在解释机器学习模型方面取得了进展，但理解生成模型中的潜在变量仍然具有挑战性。本文介绍了LatentExplainer，这是一个框架，用于自动生成深度生成模型中潜在变量的语义有意义的解释。LatentExplainer解决了三个主要挑战：推断潜在变量的含义，将解释与归纳偏差对齐，以及处理不同程度的可解释性。通过扰动潜在变量并解释生成数据的变化，该框架提供了一种系统化的方法来理解和控制数据生成过程，增强了深度生成模型的透明度和可解释性。我们在几个真实和合成数据集上评估了我们提出的方法，结果表明在生成潜在变量的高质量解释方面具有优越性。

LatentExplainer: A Framework for Generating Explanations of Latent Variables in Deep Generative Models

The paper proposes a framework called LatentExplainer that automatically generates semantically meaningful explanations of latent variables in deep generative models, enhancing the transparency and interpretability of these models.

The framework tackles three main challenges: inferring the meaning of latent variables, aligning explanations with inductive biases, and handling varying degrees of explainability. It perturbs latent variables and interprets changes in generated data to provide a systematic approach to understanding and controlling the data generation process. The proposed method is evaluated on several real-world and synthetic datasets, and the results demonstrate superior performance in generating high-quality explanations of latent variables.

Recent related work includes 'Explainable AI: A Brief Survey and Some Challenges' by Samek et al., 'Towards A Rigorous Science of Interpretable Machine Learning' by Doshi-Velez and Kim, and 'Interpretable Machine Learning: A Guide for Making Black Box Models Explainable' by Molnar.