This lecture covers the application of convex optimization in nonlinear dimensionality reduction, focusing on techniques such as the kernel trick and the spread of data points in high-dimensional spaces. The instructor explains the concept of the unfolding problem and its equivalence to an SDP, providing insights into recovering low-dimensional solutions from the SDP solution. Various examples, including piecewise constant and linear fitting, illustrate the practical implications of convex-cardinality problems in signal processing and regression tasks. The lecture concludes with discussions on exact solutions, -norm heuristics, and the interpretation of convex relaxation in addressing cardinality constraints.
This video is available exclusively on Mediaspace for a restricted audience. Please log in to MediaSpace to access it if you have the necessary permissions.
Watch on Mediaspace