Increase training set resolution, use xt-1-1, linear learning rate

Initial commit