English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

While deep neural networks take loose inspiration from neuroscience, it is anopen question how seriously to take the analogies between artificial deepnetworks and biological neuronal systems. Interestingly, recent work has shownthat deep convolutional neural networks (CNNs) trained on large-scale imagerecognition tasks can serve as strikingly good models for predicting theresponses of neurons in visual cortex to visual stimuli, suggesting thatanalogies between artificial and biological neural networks may be more thansuperficial. However, while CNNs capture key properties of the averageresponses of cortical neurons, they fail to explain other properties of theseneurons. For one, CNNs typically require large quantities of labeled input datafor training. Our own brains, in contrast, rarely have access to this kind ofsupervision, so to the extent that representations are similar between CNNs andbrains, this similarity must arise via different training paths. In addition,neurons in visual cortex produce complex time-varying responses even to staticinputs, and they dynamically tune themselves to temporal regularities in thevisual environment. We argue that these differences are clues to fundamentaldifferences between the computations performed in the brain and in deepnetworks. To begin to close the gap, here we study the emergent properties of apreviously-described recurrent generative network that is trained to predictfuture video frames in a self-supervised manner. Remarkably, the model is ableto capture a wide variety of seemingly disparate phenomena observed in visualcortex, ranging from single unit response dynamics to complex perceptual motionillusions. These results suggest potentially deep connections between recurrentpredictive neural network models and the brain, providing new leads that canenrich both fields.

A neural network trained to predict future video frames mimics critical properties of biological neuronal responses and perception