project summary
--Kernel K-means and Inverse Weighed K-Means method as a robust method for substituting traditional K-Means
--Studied and implemented the Trimmed K-means method under the Lasso and Ridge constraint to deal with irrelevant features and sparse data;
--Used the gradient descent method with a shift in the constraint boundary to get numeric solution
--Applied these robust K-Means algorithm to pattern recognition of handwritten digits
files
--The Matlab codes for the four method implemented in this project are in the corresponding subfiles
--The project description as well as the final project is in the file "Report.pdf"
--See "poster.pptx" for the poster made for the final presentation
-------The final project is a team project done by Adam Norton, Yinghe and me. I mainly did two parts, first is the inverse weighted K-Means method, and I also participated in modeling of Trimed K-means, especially I derive the numerical solution under lasso and Ridge constraint. I'm familiar with all parts of the project.