Overview
This course focuses on teaching multiscale methods for machine learning, specifically addressing the challenge of developing models that perform well for large output spaces with limited training data. The course aims to provide learners with the skills to build a multi-scale machine learning framework called Prediction for Enormous and Correlated Output Spaces (PECOS). The teaching method involves presenting a hierarchy over the output space using unsupervised learning and learning a machine learning model that makes predictions at each level of the hierarchy. The intended audience for this course includes individuals interested in machine learning, particularly those working on problems with large output spaces and limited training data.
Syllabus
Introduction
Outline
Frameworks
Linear Methods
Multioutput prediction
Textbooks
Methodology
Challenges
Weighted Graph Cuts
Weighted Kernel KMeans
Semantic Indexing
Machine Learning Training
Beam Search
Masked sparse chunk multiplication
Sparse column format
Data structure
Cache efficiency
Experimental results
Conclusion
Questions
Closing remarks
Thank you
Attendance figures
Taught by
Society for Industrial and Applied Mathematics