What are the best resources to use when starting machine learning for an experienced programmer?

Mitchell's book is very good, but it is really outdated, and maybe even slightly misleading, if looked at from a post-2000 viewpoint. It's missing SVMs, online learning, most learning theory surrounding SVM (VC theory, etc) and online learning, regularization, sparsity, most graphical model topics (LDA, for example), structured learning (CRFs, M3Ns), boosting, learning reductions, kernels, unsupervised learning as is common in NLP land, deep learning, etc; it also focuses a lot on decision trees, rules, genetic algorithms, and logical learning (most of which have found to be not so good, except maybe for MLNs for logical learning).

Also the reinforcement learning theory as presented in the book is very outdated.

Also, integrals and matrices are very useful in machine learning, and a real understand of the field cannot be achieved without a bit of linear algebra, probability, calculus, and, increasingly, optimization.

I agree with Alexandre, Mitchell's book misses too many now-relevant goodies.

Thanks for mentioning what is missing. I consider Mitchell's book to be helpful, because: 1) Experienced programmer correlates with rusty integrals. 2) I haven't seen the inductive bias explained so well elsewhere. Others also like the book.