Ask HN: Machine learning for scientists?

What's your recommendation to get up to speed on current practices in machine learning for someone ~Phd-level in the hard sciences? In other words, someone experienced in linear algebra, statistics, programming, and experiment design+evaluation, but not in the specifics of machine learning. I'm thinking courses, tutorials, blog posts, papers, books, software, instructive problem sets, etc.