方法对比

	可解释多层感知机	可解释长短期记忆网络
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2010s–present	2017–2019
提出者	Lundberg & Lee (SHAP); Ribeiro et al. (LIME); broader XAI community	Lundberg & Lee (SHAP); Ribeiro et al. (LIME); community synthesis
类型	Supervised feedforward neural network with interpretability layer	Interpretable deep learning (post-hoc explainability)
开创性文献	Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774.	Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774.
别名	XMLP, Interpretable MLP, Explainable feedforward neural network, Transparent MLP	XAI-LSTM, interpretable LSTM, LSTM with SHAP, transparent LSTM
相关	4	5
摘要	An Explainable Multilayer Perceptron (XMLP) is a standard feedforward neural network trained with backpropagation, augmented with post-hoc interpretability techniques — such as SHAP values, LIME, or integrated gradients — that attribute each prediction to individual input features. The combination retains the MLP's approximation power while satisfying transparency requirements common in regulated or high-stakes domains.	Explainable LSTM pairs a trained Long Short-Term Memory network with post-hoc interpretability techniques — chiefly SHAP, LIME, integrated gradients, or attention visualization — to reveal which time steps, tokens, or features drive each prediction. It bridges the accuracy of recurrent deep learning with the transparency demanded by high-stakes domains such as clinical decision support, fraud detection, and regulatory compliance.