Maithra Raghu
Co-founder and CEO @ Samaya AI, PhD in Machine Learning from Cornell University
About
Welcome! I am co-founder and CEO at Samaya AI, building an AI-powered knowledge-discovery platform.
From 2015 to 2022, I was a Research Scientist with the incredible Google Brain, studying representations and learning in cutting-edge deep learning systems. I received my PhD in Computer Science (Machine Learning) at Cornell University and in close collaboration with Google Brain, advised by Jon Kleinberg. I completed my undergraduate in Mathematics at the University of Cambridge (Trinity College).
I have many fond memories of competing in national mathematics Olympiads and representing the UK team in high school, where my interests in math and computer science began.
Research
My research has focused on understanding how deep neural networks learn from large scale data, the way they represent their acquired knowledge, and the effects on their capabilities. I’ve been fortunate to have many wonderful mentors on these topics, including Jon Kleinberg, Samy Bengio, Quoc Le, Geoff Hinton and Oriol Vinyals.
Writing
I like sharing thoughts on Machine Learning research and applications on Twitter and (occasionally) my blog.
news
Dec 22, 2020 | Selected Awards: Delighted to be named one of STAT’s 2020 Wunderkinds for our work on human-AI collaboration in healthcare. More details of our work are discussed in this article. I’ve also previously been named one of the Forbes 30 Under 30 in Science and the MIT Rising Stars in EECS. |
Dec 20, 2020 | Selected Talks Some of my invited talks and keynotes: Weights & Biases, RAAIS, Yale, Harvard, MIT, NYU, NVIDIA GTC 2020, NeurIPS ML for Health Workshop, O’Reilly’s AI Conference, Simons Institute Frontiers of Deep Learning, Stanford’s HealthAI Hackathon, WiML |
Sep 18, 2020 | Selected Misc: I’ve now completed my PhD at Cornell! Here is my thesis defense video My dissertation is also online. Together with Eric Schmidt, I wrote a survey overviewing many of the recent advances in deep learning, with additional pointers and advice on implementation. |
- On the Origins of the Block Structure Phenomenon in Neural Networks
- Thao Nguyen, Maithra Raghu, Simon Kornblith
- Preprint
- Pointer Value Retrieval: A new benchmark for understanding the limits of neural network generalization
- Chiyuan Zhang*, Maithra Raghu*, Jon Kleinberg, Samy Bengio
- Preprint
- Do Vision Transformers See Like Convolutional Neural Networks?
- Maithra Raghu, Thomas Unterthiner, Simon Kornblith, Chiyuan Zhang, Alexey Dosovitsky
- Advances in Neural Information Processing Systems (NeurIPS) 2021
- Teaching with Commentaries
- Aniruddh Raghu, Maithra Raghu, Simon Kornblith, David Duvenaud, Geoffrey Hinton
- International Conference on Learning Representations (ICLR) 2021
- Do Wide and Deep Neural Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth
- Thao Nguyen, Maithra Raghu, Simon Kornblith
- International Conference on Learning Representations (ICLR) 2021
- Also appeared in NeurIPS 2020 Workshop on Inductive Biases and WiML 2020
- Talk
- Anatomy of Catastrophic Forgetting: Hidden Representations and Task Semantics
- Vinay Ramasesh, Ethan Dyer, Maithra Raghu
- International Conference on Learning Representations (ICLR) 2021
- Also Best Paper ICML 2020 Workshop on Continual Learning
- Insights on Deep Representations for Machine Learning Systems and Human Collaborations
- Maithra Raghu
- PhD Thesis
- Video
- A Survey of Deep Learning for Scientific Discovery
- Maithra Raghu, Eric Schmidt
- Preprint
- Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML
- Transfusion: Understanding Transfer Learning with Applications to Medical Imaging
- Maithra Raghu*, Chiyuan Zhang*, Jon Kleinberg†, Samy Bengio† (*equal contribution) (†equal contribution)
- Neural Information Processing Systems (NeurIPS), 2019
- Also oral presentation in NeurIPS ML for Health Workshop, 2019
- Blogpost
- The Algorithmic Automation Problem: Prediction, Triage and Human Effort
- Maithra Raghu, Katy Blumer, Greg Corrado, Jon Kleinberg, Ziad Obermeyer, Sendhil Mullainathan
- Workshop on Machine Learning for Health (ML4H), NeurIPS 2018
- Direct Uncertainty Prediction for Medical Second Opinions
- Maithra Raghu*, Katy Blumer*, Rory Sayres, Ziad Obermeyer, Sendhil Mullainathan, Jon Kleinberg (*equal contribution)
- International Conference on Machine Learning (ICML), 2019
- Blogpost
- Insights on Representational Similarity in Neural Networks with Canonical Correlation
- Adversarial Spheres
- Justin Gilmer, Luke Metz, Fartash Faghri, Samuel S. Schoenholz, Maithra Raghu, Martin Wattenberg, Ian Goodfellow
- Preprint
- Appeared in International Conference on Learning Representations (ICLR) Workshop, 2018
- Can Deep Reinforcement Learning Solve Erdos-Selfridge-Spencer Games?
- Maithra Raghu, Alex Irpan, Jacob Andreas, Robert Kleinberg, Quoc V. Le, Jon Kleinberg
- International Conference on Machine Learning (ICML), 2018
- Also appeared in International Conference on Learning Representations (ICLR) Workshop, 2018
- Code
- SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability
- Explaining the Learning Dynamics of Direct Feedback Alignment
- Justin Gilmer, Colin Raffel, Samuel S. Schoenholz, Maithra Raghu, Jascha Sohl-Dickstein
- Appeared in International Conference on Learning Representations (ICLR) Workshop, 2017
- On the Expressive Power of Deep Neural Networks
- Maithra Raghu, Ben Poole, Jon Kleinberg, Surya Ganguli, Jascha Sohl-Dickstein
- International Conference on Machine Learning (ICML), 2017
- Video
- Linear Additive Markov Processes
- Ravi Kumar, Maithra Raghu, Tamas Sarlos, Andrew Tomkins (alphabetical order)
- World Wide Web Conference (WWW), 2017
- Exponential expressivity in deep neural networks through transient chaos
- Ben Poole, Subhaneil Lahiri, Maithra Raghu, Jascha Sohl-Dickstein, Surya Ganguli
- Advances in Neural Information Processing Systems (NeurIPS) 2016
- Team Performance with Test Scores
- Jon Kleinberg, Maithra Raghu (alphabetical order)
- Economics and Computation (EC) 2015
- Invited to special issue of ACM Transactions on Economics and Computation, 2018