Translate to multiple languages

Subscribe to my Email updates
Enjoy what you've read, make sure you subscribe to my Email Updates

Sunday, March 26, 2017

What Is The Best Way To Learn Machine Learning Without Taking Any Online Courses? | Forbes

"What is the best way to start learning machine learning and deep learning without taking any online courses? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world." Quora, Contributor.

Answer by Eric Jang, Research engineer at Google Brain, on Quora:

Photo: Shutterstock

Let me first start off by saying that there is no single “best way” to learn machine learning, and you should find a system that works well for you. Some people prefer the structure of courses, others like reading books at their own pace, and some want to dive right into code.

I started with Andrew Ng’s Machine Learning Coursera course in 2012, knowing almost zero linear algebra and nothing about statistics or machine learning. Note that although the class covered neural networks, it was not a course on Deep Learning. I really enjoyed how the course formulated “machine learning” as nothing more than numerical optimization.

Deep Learning book
If online courses are too slow for you, the best consolidated resource is probably Deep Learning book by Goodfellow, Bengio, and Courville. It has a few chapters dedicated to the basics (sort of like what is covered in Ng’s class) and then jumps into practical DNNs.

Murphy’s Probabilistic
Machine Learning textbook
A statistical/mathematically rigorous background is not required to do useful Deep Learning work, but it really helps to formulate hypotheses about why models are/are not working, and what might help. Murphy’s Probabilistic Machine Learning textbook is a great foundation for mathematically rigorous ML (and has great diagrams too!) 

After you finish the DL book, you can “specialize” into one of the subfields/sub-subfields of Deep Learning, by implementing some of the papers yourself. Some example topics:
  • Bayesian Deep Learning (combining neural nets with graphical models)
  • Deep Reinforcement Learning (AlphaGo, Atari-playing AI, Robotics)
  • Generative Models (GANs, PixelCNN, VAEs)
  • Adversarial Methods (GANs, Actor-Critic)
  • Theory of Deep Learning
  • Computer Vision
  • NLP/Speech (translation, captioning, seq2seq models)
  • Symbolic reasoning (e.g. proof-solving)
  • Recurrent Neural Networks (e.g. LSTMs, external memory, attention)
  • Applications (solving domain-specific problems like classifying cancer, protein folding, lip reading from video)
  • Meta-learning / learning-to-learn (Synthetic Gradients, Pathnet)
The Deep Learning field has dramatically expanded in the last few years, to the point where it’s not realistic to grok all the subfields of Deep Learning in a short amount of time.
Read more... 

Source: Forbes