7. Rule-Based Machine Learning Algorithms

This video is Part 7 of the series "Machine Learning Essentials for Biomedical Data Science" covering the key essentials for using machine learning as part of a data science analysis pipeline. While topics are primarily framed around applications in biomedicine, this content is broadly applicable to other domains. The material presented was assembled based on my 15 years experience as a machine learning researcher and educator. I'm currently an Assistant Professor (Pending) of Computational Biomedicine at the Cedars Sinai Medical Center. Some of the slide content is original, with much adapted from a wide variety of textbooks, slides, and lectures by various authors and speakers. This video series expands upon a full-day workshop prepared for and presented at the Cedars Sinai Medical Center in Los Angeles. This video represents my own understanding and perspectives. Weblinks: http://ryanurbanowicz.com/ https://github.com/UrbsLab https://github.com/UrbsLab/STREAMLINE Chapters: 0:00 Introduction 1:35 Rule-based Machine Learning 4:07 Learning Classifier Systems (LCS) 8:14 Disadvantages 9:55 Overview of LCS Algorithm 16:25 Training Data / Instances 17:38 Rule Population / Rules 19:33 Matching 20:48 Rule vs. Classifier 23:01 Training Cycle (Review) 23:37 Covering (Rule Discovery) 24:41 Rule Parameter Updates 25:03 Implicit Rule Generalization Pressure 25:51 Subsumption 27:55 Genetic Algorithm (Rule Discovery) 29:24 Rule Deletion 30:11 Rule Compaction 30:28 Prediction Array (Making Predictions) 30:55 Learning more about LCS 31:36 Conclusion