Theory and Practice of Infinitely Wide Neural Networks - Guest Talk
- ๐ค Speaker: Roman Novak, Google Brain
- ๐ Date & Time: Wednesday 28 September 2022, 11:00 - 12:30
- ๐ Venue: Cambridge University Engineering Department, CBL Seminar room BE4-38
Abstract
A common observation that wider (in the number of hidden units/channels/attention heads) neural networks perform better motivates studying them in the infinite-width limit.
Remarkably, infinitely wide networks can be easily described in closed form as Gaussian processes (GPs), at initialization, during, and after trainingโbe it gradient-based, or fully Bayesian training. This provides closed-form test set predictions and uncertainties from an infinitely wide network without ever instantiating it (!).
These infinitely wide networks have become powerful models in their own right, establishing several SOTA results, and are used in applications including hyper-parameter selection, neural architecture search, meta learning, active learning, and dataset distillation.
The talk will provide a high-level overview of our work at Google Brain on infinite-width networks. In the first part I will derive core results, providing intuition for why infinite-width networks are GPs. In the second part I will discuss challenges and solutions to implementing and scaling up these GPs. In the third part, I will conclude with example applications made possible with infinite width networks.
The talk does not assume familiarity with the topic beyond general ML background.
Series This talk is part of the Machine Learning Reading Group @ CUED series.
Included in Lists
- All Talks (aka the CURE list)
- bld31
- Cambridge Centre for Data-Driven Discovery (C2D3)
- Cambridge Forum of Science and Humanities
- Cambridge Language Sciences
- Cambridge talks
- Cambridge University Engineering Department, CBL Seminar room BE4-38
- Cambridge University Engineering Department Talks
- Centre for Smart Infrastructure & Construction
- Chris Davis' list
- Computational Continuum Mechanics Group Seminars
- custom
- Featured lists
- Guy Emerson's list
- Hanchen DaDaDash
- Inference Group Journal Clubs
- Inference Group Summary
- Information Engineering Division seminar list
- Interested Talks
- Machine Learning Reading Group
- Machine Learning Reading Group @ CUED
- Machine Learning Summary
- ML
- ndk22's list
- ob366-ai4er
- Quantum Matter Journal Club
- Required lists for MLG
- rp587
- School of Technology
- Simon Baker's List
- TQS Journal Clubs
- Trust & Technology Initiative - interesting events
- yk373's list
- yk449
Note: Ex-directory lists are not shown.
![[Talks.cam]](/static/images/talkslogosmall.gif)

Roman Novak, Google Brain
Wednesday 28 September 2022, 11:00-12:30