classify mnist datasets using ridge regression, optimize the algorithem with SGD, stochastic dual coordinate ascent, and mini-batching
- get data from http://yann.lecun.com/exdb/mnist/,
- Reduce the 784 dimension data to 50 using PCA
- generate Random Fourier Features to approximate the RBF Kernel
- optimize ridge regression using SGD
- optimize ridge regression using stochastic dual coordinate ascent(SDCA)
- mini-batching SDCA