1985–, Computer scientist
Ian Jonathan Goodfellow is an American computer scientist whose 2014 paper Generative Adversarial Networks (with Pouget-Abadie, Mirza, Xu, Warde-Farley, Ozair, Courville and Bengio) introduced GANs: a generative-modelling framework in which a generator network learns to produce samples that fool a discriminator network trained to distinguish samples from real data. The adversarial dynamic, both networks trained simultaneously, drives the generator to produce increasingly realistic samples.
GANs reshaped generative modelling for the next decade. Variants (DCGAN, conditional GAN, CycleGAN, StyleGAN, BigGAN, Progressive GAN) enabled photorealistic image generation, style transfer, domain adaptation, super-resolution and data augmentation. StyleGAN-3 (2021) was the high-water mark of GAN-based image generation before diffusion models took over the lead in 2022.
Goodfellow co-authored the standard Deep Learning textbook (with Bengio and Courville, 2016) and has worked at Google Brain, OpenAI, Apple and DeepMind. He has been a leading voice on adversarial examples and machine-learning security throughout his career.
Video
Related people: Yoshua Bengio
Works cited in this book:
- Intriguing properties of neural networks (2013) (with Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Rob Fergus)
- Generative Adversarial Nets (2014) (with Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio)
- Explaining and Harnessing Adversarial Examples (2015) (with Jonathon Shlens, Christian Szegedy)
- Deep Learning (2016) (with Yoshua Bengio, Aaron Courville)
- Deep Learning with Differential Privacy (2016) (with Martin Abadi, Andy Chu, H. Brendan McMahan, Ilya Mironov, Kunal Talwar, Li Zhang)
Discussed in:
- Chapter 14: Generative Models, Generative Models