This is an online version of the Manning book Probabilistic Deep Learning: With Python, Keras and TensorFlow Probability MEAP V06. With liveBook you can access Manning books in-browser — anytime, anywhere.
Thanks for purchasing the MEAP of Probabilistic Deep Learning: With Python, Keras and TensorFlow Probability. If you wonder what is behind the breakthroughs of deep learning (DL), how you can build and tune highly performant DL models yourself, and what the beauty is of probabilistic models, you are the reader we have in mind. To get the most benefit from the book, it would be great if you have some first experience in machine learning or deep learning. And make sure that you understand chapter 2. It’s written to get everyone on the same page. If you understand that chapter, you’re fine for the rest of the book.
We have entered the realm of deep learning as applied statisticians fascinated by its high performance on complex data like images. When we first started to use deep learning five years ago, we had several deja vus. That so-called categorical cross-entropy—isn’t that just the same as the maximum likelihood principle? Isn’t deep learning nothing more than a complex form of probabilistic modeling after all? Then in 2018, Judea Pearl hit the nail on the head with his now famous quote, “All the impressive achievements of deep learning amount to just curve fitting.”