Convergent Temperature Representations in Artificial and Biological Neural Networks

Area of Science:

• Computational Neuroscience
• Artificial Intelligence
• Neurobiology

Background:

• Recent discoveries highlight parallels between biological neural networks (BNNs) and artificial neural networks (ANNs).
• The extent to which ANNs can elucidate BNN function remains an open question.

Purpose of the Study:

• To investigate if ANNs can provide insights into BNN function by designing an ANN for heat gradient navigation.
• To explore computational and representational similarities between the designed ANN and a known zebrafish BNN.

Main Methods:

• Designed and trained an artificial neural network (ANN) to perform heat gradient navigation.
• Compared the ANN's computational strategies and heat representations with those of a biological neural network (BNN) in zebrafish.
• Constrained the ANN with the C. elegans motor repertoire to test generalization.

Main Results:

• The ANN exhibited striking computational and heat representation similarities to the zebrafish BNN, including shared ON- and OFF-type representations.
• The ANN's function depended on zebrafish-like units, and its accessibility led to the discovery of a novel temperature-responsive cell type in the zebrafish cerebellum.
• Constraining the ANN with C. elegans motor data altered sensory representations, indicating the approach's generalizability.

Conclusions:

• Artificial neural networks (ANNs) and biological neural networks (BNNs) converge on stereotypical representations.
• ANNs serve as a powerful tool for understanding the function of their biological counterparts.
• This study successfully utilized ANNs to uncover novel biological insights.