Abstract

Please join us for our fortnightly journal club online via zoom where two presenters will jointly present a topic together. The next topic is ‘Deep neural networks as models for the visual cortex’ presented by Yashar Ahmadian and Gido van de Ven.

Zoom information: https://us02web.zoom.us/j/84958321096?pwd=dFpsYnpJYWVNeHlJbEFKbW1OTzFiQT09 Meeting ID: 841 9788 6178 Passcode: 659046

Summary: In recent years, deep neural networks (DNNs) have enjoyed considerable success as computational models for the brain’s ventral visual stream. After a short introduction to this field, for which we follow the review by Yamins & DiCarlo (2016), we discuss the paper by Bashivan et al. (2019) in part 1 of this journal club. This paper used a DNN , trained on Imagenet in a supervised fashion, as a model of the visual stream to synthesize images predicted to selectively activate certain neurons in the macaque visual cortex, which they then tested by presenting the synthesized images back to the macaque. In part 2, we discuss the recent paper by Zhuang et al. (2021), which asks whether training DNNs in an unsupervised fashion, rather than in the typical supervised fashion, can produce better models of the visual stream.

Relevant reading:

Yamins, D. L., & DiCarlo, J. J. (2016). Using goal-driven deep learning models to understand sensory cortex. Nature neuroscience, 19(3), 356. https://www.nature.com/articles/nn.4244

Bashivan, P., Kar, K., & DiCarlo, J. J. (2019). Neural population control via deep image synthesis. Science, 364(6439). https://science.sciencemag.org/content/364/6439/eaav9436.abstract

Zhuang, C., Yan, S., Nayebi, A., Schrimpf, M., Frank, M. C., DiCarlo, J. J., & Yamins, D. L. (2021). Unsupervised neural network models of the ventral visual stream. Proceedings of the National Academy of Sciences, 118(3). https://www.pnas.org/content/118/3/e2014196118.short