Generative Adversarial Networks (GANs) by Udacity – The complete YouTube playlist

Christos - Iraklis Tsatsoulis Data Science, Deep Learning 1 Comment

Have you heard of Generative Adversarial Networks (GANs)? If you are in the machine learning & data science business, you should – it has been argued that GANs will change the world, and Yann LeCun, one of the pioneers of modern deep learning and currently Director of Artificial Intelligence Research at Facebook, has called them “the coolest idea in deep learning in the last 20 years”.

GANs are the coolest idea in deep learning in the last 20 years

Yann LeCun, Director, Facebook AI Research

Wired magazine recently published an enjoyable piece on the conception of GANs by Ian Goodfellow, inventor of GANs and one of “the world’s most important AI researchers“.  So, it was a very pleasant surprise to find out that Goodfellow himself gives the GANs introductory lectures in the recently launched Deep Learning Foundations Nanodegree by Udacity.

Nevertheless, and despite the fact that the GANs unit lectures are available at YouTube, there are not grouped together in a playlist, making it close to impossible for people not enrolled in the Nanodegree to locate them. So, as a service to the community, we hereby provide links for the complete GANs unit videos at YouTube (duration times are in min:sec, videos by Ian Goodfellow are indicated with an asterisk):

Intro to GANs

  1. Introducing Ian Goodfellow (0:47)*
  2. What can you do with GANs? (4:22)*
  3. How GANs work (4:46)*
  4. Games and Equilibria (7:16)*
  5. Practical tips and tricks for training GANs (7:54)*
  6. Get started with a GAN (2:51)
  7. Generator Network (4:13)
  8. Discriminator Network (0:39)
  9. Generator and Discriminator Solutions (3:42)
  10. Building the Network (2:21)
  11. Building the Network Solution (2:15)
  12. Training Losses (2:25)
  13. Training Optimizers (1:51)
  14. Training Losses and Optimizers Solution (3:15)
  15. A Trained GAN (6:23)

Jupyter notebooks for the above lectures are available here.

Deep Convolutional GANs

  1. Deep Convolutional GAN (DCGAN) architecture (4:14)
  2. DCGAN and the Generator (7:01)
  3. Generator solution (2:38)
  4. Discriminator (1:52)
  5. Discriminator solution (2:09)
  6. Building and training the network (3:33)
  7. Hyperparameters solution (2:45)

Jupyter notebooks for the above lectures are available here. There are also some extra notebooks on batch normalization.

Semi-Supervised learning with GANs

  1. Semi-Supervised Classification with GANs (5:56)*
  2. Introducing Semi-Supervised Learning (3:07)*
  3. Data prep (3:17)*
  4. Building the Generator and the Discriminator (9:53)*
  5. Model loss exercise (8:08)*
  6. Model optimization exercise (6:50)*
  7. Training the network (3:05)*
  8. Discriminator solution (3:47)*
  9. Model loss solution (4:34)*
  10. Model optimizer solution (2:11)*
  11. Trained Semi-Supervised GAN (3:55)*

Jupyter notebooks for the above lectures are available here.

Many thanks to Udacity for making the videos and notebooks publicly available; enjoy and happy coding!-

Christos - Iraklis Tsatsoulis

Christos - Iraklis is one of our resident Data Scientists. He holds advanced graduate degrees in applied mathematics, engineering, and computing. He has been awarded both Chartered Engineer and Chartered Manager status in the UK, as well as Master status in Kaggle.com due to "consistent and stellar results" in predictive analytics contests.

Leave a Reply

1 Comment on "Generative Adversarial Networks (GANs) by Udacity – The complete YouTube playlist"

Notify of
avatar
Sort by:   newest | oldest | most voted
JC
Guest

Thank you! Quite useful list 🙂

wpDiscuz