As a visual learner, I personally think that Ng's coursera is the best intro. Having the schedule and assignments also helps motivate you to finish it.
If you're interested in getting a better idea of what is involved in ML, I recommend checking out the blogposts on kaggle written by the competition winners—
http://blog.kaggle.com/category/winners-interviews/ I forget which article it was that I liked a lot, but it was basically two statisticians working at healthcare companies that won, and talked about what led them to choose their model, and how it works.
If you want to take the SICP-esque approach, the bible of ML is ESLR (not in a meme-sense like SICP), but I recommend starting with ISLR. Yes, it takes 30 more pages to reach linear regression, and doesn't cover neural nets as in-depth as its older brother does (nobody reads ESLR for DNNs anyway), but the math is a better pace. Plus, as someone who primarily uses Python, it was neat to see how R works.
It's free here:
http://www-bcf.usc.edu/~gareth/ISL/
