Learn about data influence and its critical role in data science through this comprehensive lecture that explores influence functions, efficiency, and convexity. Delve into the mathematical foundations of data influence, starting with core motivations and progressing through detailed explanations of influence functions. Examine efficiency considerations and convexity principles, with special attention to IF-LLO misalignment and PBRF concepts. Master key mathematical concepts including the relationship between strictly positive definite matrices, invertibility, and the existence of minimum values in convex functions. Follow along with clearly structured segments covering theoretical foundations and practical applications, concluding with a thorough recap that reinforces essential concepts.
Overview
Syllabus
Lecture starts
Data influence motivation
Influence functions
Efficiency
Convexity
IF-LLO misalignment
PBRF
Recap
Taught by
UofU Data Science