Phillip Isola
Professor · MIT
Phillip Isola is a professor at MIT working on deep learning, computer vision and generative models, known for conditional adversarial image-to-image translation and CycleGAN.
- lectures
- 10
- series
- 1
- hours
- 13
Start here
Series and courses
MIT 6.7960: Deep Learning
In these lectures
- Graph neural networks are generalization of convolutional networks designed to operate on arbitrary graph-structured data.Architectures: Graphs
- Always inspect data format, shape, min, max, and variance right before the forward pass.Hacker's Guide to Deep Learning
- Structured prediction problems require modeling high-dimensional data where output dimensions exhibit strong joint dependencies rather than statistical independence.Generative Models: Conditional Models
Draws on
People cited in the lectures, at the moment of the citation.
- Dylan Hadfield-MenellSlide adaptation source
- James GlassGlass, 2007, cited for automatic speech recognition.
- Andrew NgCited for his popular slide illustrating deep learning scaling with data.
- Sanjeev AroraCited for papers and talks discussing theoretical aspects of transfer learning.
- Yann LeCunCited for the intelligence cake analogy describing information given during learning.
- Noam ChomskyCited for the sentence 'Colorless green ideas sleep furiously'.
Playlists
1About
Phillip Isola is a professor at MIT working on deep learning, computer vision and generative models, known for conditional adversarial image-to-image translation and CycleGAN. His MIT 6.7960 lectures build transformers and graph neural networks, derive the VAE evidence bound and read scaling laws.
Taught with
Same subject, AI & Machine Learning
Connections
16Their places, courses, books and lectures.






