The following are the takeaways from Ian GoodFellow Interview.

  • Inventor of GAN
  • Coded up GAN in one night
  • Learn the basic math needed for Deep Learning - Linear Algebra + Probability
  • Wrote a book with PhD co-advisors
  • GAN can be useful in many different fields
  • Got kicked after using Deep Belief networks from Geoff Hinton
  • Do you want to make unsupervised work same as deep learning algos
  • Do you want to make reinforcement learning work same as deep learning algos
  • Do you want to work on the bias ?
  • There are ton of ways you can contribute to AI ?
  • PhD is not required
  • Put your code on github
  • Writing papers and put in on arxiv is also important
  • Always choose a project to work with - Along with the learnings from the book
  • Exercise basic skills - Always work on project
  • Adverserial examples - Machine Learning security
  • Spends 40% of the time on improving the stability of GANs
  • Ian Good Fellow’s book is good - All the needed math is there in the first few chapters of the book

Obviously the first thing that comes to my mind - Why have I not being working on Deep Learning book ? Why am I shying away from going through the math and working out stuff ?

I should also learning about GAN’s and application in finance? There are a ton of applications of GAN. Need to be creating a demo using GAN