Computational architecture of the octopus visual system

Abstract

Our understanding of the neural basis of complex visual function is largely based on studies in vertebrates, from fish to primates, motivated by their similarity to the human brain. Cephalopods, including octopuses, represent a completely independent evolution of intelligent behavior and a highly capable visual system, but with a fundamentally different brain organization and neural architecture. Indeed, the octopus brain is so radically different from vertebrates that they are often described in the scientific literature as alien. Understanding cephalopod vision would therefore dramatically expand our knowledge of sensory processing and cognition, by extending it beyond vertebrates to include a second instantiation of a highly capable visual processing system, and could lead to transformative solutions in brain-inspired computing. However, the neural basis of visual processing in cephalopods is almost completely unknown.In this proposal, our goal is to determine how neurons in the octopus visual system encode information and perform the computations necessary for vision. We will first measure neural activity in large populations of neurons in the visual center of the octopus brain, using two-photon calcium imaging while presenting visual stimuli, to determine how neurons process visual information. We will then map the wiring of the visual center to the rest of the brain, by performing structural tracing with retrograde tracers, which will allow us to link functional properties of neurons with specific behavioral capabilities. Finally, we will implement these findings into computational models for machine vision, by determining the information encoded in specific output pathways, and by developing experimentally-constrained neural network models. Much as studies in the vertebrate visual system have inspired current approaches to machine vision and AI, such as deep neural networks, we expect that these findings will contribute to novel approaches based on the unique computational architecture of the octopus brain.

Document Details

Document Type
DoD Grant Award
Publication Date
May 05, 2021
Source ID
N000142112426

Entities

People

  • Cristopher M Niell

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Oregon

Tags

Fields of Study

  • Biology
  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Neural Network Machine Learning.
  • Neuroscience

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks