报告题目：Advancing Green Machine Learning: A Perspective of Granular Computing

报告所属学科：管理科学与工程

报告人：Witold Pedrycz（加拿大阿尔伯塔大学）

报告时间：2022年9月14日、2022年9月21日、2022年9月28日、

2022年10月12日，上午9：00-10：30

报告地点：腾讯会议：415-8231-4482  QQ群：687807963

报告摘要：

(1) Introductory comments: Fundamentals and Challenges

The key agenda of Machine Learning. Main concepts. Challenges of Machine Learning: Interpretability, coping with numerous data sources, curse of dimensionality, agenda of green machine learning.

(2) Unsupervised learning

The concept of unsupervised learning. Classes of methods, interpretation, Information granules. From information granules to rule-based computing.

(3) Dimensionality reduction and combining classifiers

Concentration effect. Principal component analysis and random projection. Autoencoder, nonnegative matrix factorization, Salmon stress function. Relational factorization. Interpretability of results. Combination of classifiers: voting, bagging, boosting, mixture of experts.

(4) Federated Learning and Transfer Learning

Motivating factors behind federated learning: coping with data islands, average and gradient federated learning, Federated learning-based rule design, performance analysis. Transfer learning and knowledge reuse.  Green Machine Learning. Design of granular models.  Knowledge distillation: teacher - student paradigm in model compression. Granular regularization.

报告人简介：

Witold Pedrycz，加拿大阿尔伯塔大学讲席教授，加拿大工程院院士，加拿大皇家科学院院士，波兰科学院外籍院士，电气和电子工程师协会会士。现担任《Information Sciences》和《WIREs Data Mining and Knowledge Discovery》主编，以及《Int. J. of Granular Computing》和《J. of Data Information and Management》共同主编。主要研究方向包括智能计算、模糊建模和颗粒计算、知识发现和数据科学、模式识别、基于知识的神经网络和控制工程等；并发表多篇高质量论文，出版著作21本，H指数119，计算机科学和电子学排名第82位。