I strongly believe machine learning methods should not be tied to programming languages. If tomorrow a new language comes out that offers better performance than Go, while keeping the developer friendliness of Go, I'd move to that language in a heartbeat. I wouldn't have to be worried about having to relearn new machine learning methods. I already know them. I can simply rewrite them in that new language. As such, my recommendations would be language-agnostic.
If you want to learn more about machine learning algorithms, I recommend Christopher Bishop's, Pattern Recognition and Machine Learning. It's a slightly older book, but you'll be surprised at how many new developments in machine learning have their roots in that tome.
If you want to learn more about deep learning, I recommend Ian Goodfellow and Yoshua Bengio's, Deep Learning...
