University of Cambridge > Talks.cam > Computational Neuroscience > Certain about uncertainty? Latent representations of VAEs optimized for visual tasks.

  Browse A-Z

Log in

Cambridge users (raven)details Other usersdetails No account?details

Information on

Subscribing to talksdetails Finding a talkdetails Adding a talkdetails Disseminating talksdetails Help and Documentationdetails

Certain about uncertainty? Latent representations of VAEs optimized for visual tasks.

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Daniel Kornai.

Deep Learning methods are increasingly becoming instrumental as modeling tools in Computational Neuroscience, employing optimality principles to build bridges between neural responses and perception or behavior. Deep Generative Models (DGMs) can learn flexible latent variable representations of images while avoiding intractable computations, common in Bayesian inference. However, investigating the properties of inference in Variational Autoencoders (VAEs), a major class of DGMs, reveals severe problems in their uncertainty representations. Here we draw inspiration from classical computer vision to introduce an inductive bias into the VAE by incorporating a global explaining-away latent variable, which remedies defective inference in VAEs. Unlike standard VAEs, the Explaining-Away VAE (EA-VAE) provides uncertainty estimates that align with normative requirements across a wide spectrum of perceptual tasks, including image corruption, interpolation, and out-of-distribution detection. We find that restored inference capabilities are delivered by developing a motif in the inference network (the encoder) which is widespread in biological neural networks: divisive normalization. Our results establish EA-VAEs as reliable tools to perform inference under deep generative models with appropriate estimates of uncertainty.

This talk is part of the Computational Neuroscience series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

Other lists

Considering Performance: A Symposium of American Culture and Literature CMS Events ReproSoc

Other talks

Playing with Fire - Parties and Political Violence in Kenya and India Are animals aware of their own mortality? Predicting recurrence of prostate cancer: a Bayesian approach Acoustics of musical instruments - why is a saxophone like a violin? XXIV Old Buildings Lecture 2025: 'Equitable Ownership' The Book of protecting Osiris among those of the Netherworld: An annotated version of the First Hour.
 

© 2006-2025 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity