RI: Small: Enabling Interpretable AI via Bayesian Deep Learning

RI：小型：通过贝叶斯深度学习实现可解释的人工智能

• 批准号: 2127918
• 负责人: Hao Wang
• 金额: $ 49.99万
• 依托单位: Rutgers University New Brunswick
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2127918&HistoricalAwards=false
• 关键词: RI; Small; Enabling; Interpretable; AI

Interpretability is one of the fundamental obstacles on the adoption and deployment of deep-learning-based AI systems across various fields such as healthcare, e-commerce, transportation, earth science, and manufacturing. An ideal interpretable model should be able to interpret its prediction using human-understandable concepts (e.g., “color” and “shape”), conform to conditional dependencies in the real world (e.g., whether a customer's purchase is due to a discount), and handle uncertainty in data (e.g., how certain the model is about the rainfall tomorrow). Unfortunately, deep learning as a connectionist approach does not natively support these desiderata. The goal of this project is to develop a general interpreter framework for deep learning models. Interpreters under this framework can be plugged into a deep learning model and interpret its predictions using a graph of human-understandable concepts, without sacrificing the model’s performance. Methods developed in this project will be applied in health monitoring to interpret models’ reasoning on patient status, and in recommender systems to interpret models’ recommended items for users.This project will develop two sets of methods based on Bayesian deep learning: (1) “Bayesian deep interpreters” that interpret deep learning models with graphical models describing the conditional dependencies leading to current predictions. (2) “Bayesian deep controllers” that control deep learning models' predictions by manipulating specific random variables in the graphical models attached to the controlled models. Development of such novel methods will build intellectual and formal connection between deep learning and probabilistic graphical models, two major machine learning paradigms that have long been seen as incompatible. It will advance the state of the art on machine learning and AI by: (1) formulating a new Bayesian deep learning framework to unify deep learning and graphical models, the synergy of which will significantly improve deep learning interpretability, (2) under such a principled framework, designing concrete methods that are plug-and-play and therefore do not sacrifice the deep learning models' performance (e.g., accuracy), (3) investigating what theoretical guarantees the developed methods provide and therefore laying foundations for future work by the team and the community, (4) analyzing the trade-off between accuracy, interpretability, and controllability and providing design guidance for interpretable AI systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

可解释性是在医疗保健、电子商务、交通、地球科学和制造等各个领域采用和部署基于深度学习的人工智能系统的基本障碍之一。理想的可解释模型应该能够解释其预测。使用人类可理解的概念（例如“颜色”和“形状”），符合现实世界中的条件依赖关系（例如，客户的购买是否是由于折扣），并处理数据的不确定性（例如，不幸的是，深度学习作为联结主义方法本身并不支持这些需求。该项目的目标是开发一个可以插入该框架下的深度学习模型的通用解释器框架。进入深度学习模型并使用人类可理解的概念图来解释其预测，而不牺牲模型的性能。该项目中开发的方法将应用于健康监测以解释模型对患者状态的推理，并应用于推荐系统来解释。该项目将开发两套基于贝叶斯深度学习的方法：（1）“贝叶斯深度解释器”，用描述导致当前预测的条件依赖性的图形模型解释深度学习模型。贝叶斯深度控制器通过操纵附加到受控模型的图形模型中的特定随机变量来控制深度学习模型的预测，这种新颖方法的开发将在深度学习和概率图形模型这两种主要机器之间建立智能和形式的联系。它将通过以下方式推动机器学习和人工智能的发展：（1）制定新的贝叶斯深度学习框架来统一深度学习和图模型，其协同作用将显着提高。深度学习的可解释性，（2）在这样的原则框架下，设计即插即用的具体方法，因此不会牺牲深度学习模型的性能（例如准确性），（3）研究什么理论保证了所开发的方法提供并因此为团队和社区未来的工作奠定基础，（4）反映分析准确性、可解释性和可控性之间的权衡，并为可解释的人工智能系统提供设计指导。该奖项授予 NSF 的法定使命，并被认为值得通过以下方式支持：使用基金会的智力价值和更广泛的影响审查标准进行评估。

项目成果

Extrapolative continuous-time Bayesian neural network for fast training-free test-time adaptation

用于快速免训练测试时间适应的外推连续时间贝叶斯神经网络

• 发表时间: 2022-12
• 期刊: Thirty-Sixth Annual Conference on Neural Information Processing Systems (NeurIPS
• 作者: Huang, Hengguan;Gu, Xiangming;Wang, Hao;Xiao, Chang;Liu, Hongfu;Wang, Ye
• 通讯作者: Wang, Ye

Fairness in Recommendation: Foundations, Methods and Applications

推荐的公平性：基础、方法和应用

• DOI: 10.1145/3610302
• 发表时间: 2023-07
• 期刊: ACM Transactions on Intelligent Systems and Technology
• 影响因子: 5
• 作者: Li, Yunqi;Chen, Hanxiong;Xu, Shuyuan;Ge, Yingqiang;Tan, Juntao;Liu, Shuchang;Zhang, Yongfeng
• 通讯作者: Zhang, Yongfeng

Learn Basic Skills and Reuse: Modularized Adaptive Neural Architecture Search (MANAS)

学习基本技能和重用：模块化自适应神经架构搜索（MANAS）

• DOI: 10.1145/3511808.3557385
• 发表时间: 2022-10
• 期刊: CIKM '22: Proceedings of the 31st ACM International Conference on Information & Knowledge Management
• 作者: Chen, Hanxiong;Li, Yunqi;Zhu, He;Zhang, Yongfeng
• 通讯作者: Zhang, Yongfeng

User-Controllable Recommendation via Counterfactual Retrospective and Prospective Explanations

通过反事实的回顾性和前瞻性解释进行用户可控的推荐

• 发表时间: 2023-09
• 期刊: Proceedings of the 26th European Conference on Artificial Intelligence
• 作者: Juntao Tan; Yingqiang Ge
• 通讯作者: Yingqiang Ge

ExplainableFold: Understanding AlphaFold Prediction with Explainable AI

ExpandableFold：通过可解释的 AI 理解 AlphaFold 预测

• DOI: 10.1145/3580305.3599337
• 发表时间: 2023-08
• 期刊: KDD '23: Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
• 作者: Tan, Juntao;Zhang, Yongfeng
• 通讯作者: Zhang, Yongfeng