Dimensionality Reduction and Manifold Estimation PMF-Mathematics Winter, 2022 Project Suggestions These projects should be done individually, without collaboration. Prepare a talk with approximately 20 slides, expecting to take 30 minutes plus 15 minutes for questions. NOTE: The references are not exhaustive: you are welcome to find other sources. 1. Prove some version of the Whitney Embedding Theorem, for smooth noncompact manifolds, or for compact differentiable manifolds. Discuss the technical issues and the important differences in the various versions. References: https://www.math.wustl.edu/~victor/classes/pmf/WOMPtalk-Manifolds.pdf https://www.math.wustl.edu/~victor/classes/pmf/PartOfUnityLocFiniteRefinements.pdf https://www.math.wustl.edu/~victor/classes/pmf/Brahim-Abdenbi-presentation-Whitney.pdf https://www.math.wustl.edu/~victor/classes/pmf/WhitEmb-Lec09.pdf https://www.math.wustl.edu/~victor/classes/pmf/lecture10-WhitEmbed.pdf 2. Prove the Birkhoff Ergodic Theorem. Prove the Von Neumann Ergodic Theorem. Discuss their differences and compare them with the mean and maximal ergodic theorems. Try to find hypotheses that guarantee a given rate of convergence, and find examples with particularly slow convergence. References: https://www.math.wustl.edu/~victor/classes/pmf/Yunis.pdf https://www.math.wustl.edu/~victor/classes/pmf/MetropHastingsEtc.pdf 3. Find coordinates in 2 dimensions for the full 2000-point Swiss Roll data set using graph Laplacian eigenvectors (for an undirected unweighted adjacency matrix with some reasonable choice of edges) and diffusion maps (for a weighted similarity matrix with some choice of sigma). Experiment with parameters and discuss how this changes the results. References: https://www.math.wustl.edu/~victor/classes/pmf/sr2000x3.dat https://www.math.wustl.edu/~victor/classes/pmf/06swiss.txt https://www.math.wustl.edu/~victor/classes/pmf/ch1-EigGraphLapl.pdf https://www.math.wustl.edu/~victor/classes/pmf/GraphLaplacian-tutorial.pdf https://www.math.wustl.edu/~victor/classes/pmf/Lafon06.pdf https://www.math.wustl.edu/~victor/classes/pmf/delaPorte-Herbst-Hereman-vanderWalt-PRASA-2008.pdf