How amolgm/ufldl_selftaughtlearning works
Project type and purpose
• This is a MATLAB-based machine-learning exercise (“SelfTaughtLearning”) implementing unsupervised feature learning and supervised classification on the MNIST dataset.
File structure
• Top level: README.md and a single folder stl_exercise.
• stl_exercise/code: all MATLAB .m files plus the minFunc subfolder. This is your core ML pipeline and optimizer library.
• stl_exercise/code/minFunc: a self-contained optimization toolkit (gradient, Hessian, line search routines) with both MATLAB and MEX/C implementations.
• stl_exercise/data/mnist: raw MNIST files (images and labels, both compressed and uncompressed).
README for additional insights
• The README confirms this is Andrew Ng’s “Self-taught learning” exercise.