Broken time-reversal symmetry in visual motion detection

Area of Science:

• Neuroscience
• Computational Neuroscience
• Vision Science

Background:

• Classical models of motion estimation assume perfect time reversal symmetry.
• Biological visual systems are expected to mirror this symmetry in motion perception.

Purpose of the Study:

• To investigate whether biological motion perception exhibits time reversal symmetry.
• To identify stimuli and conditions that reveal symmetry breaking in visual systems.
• To explore the role of neural network properties in time reversal symmetry breaking.

Main Methods:

• Designed specific visual stimuli to test time reversal symmetry in fruit fly (Drosophila) optomotor behavior.
• Trained neural network models to predict scene velocity using naturalistic and artificial contrast distributions.
• Analyzed model responses analytically and numerically to identify sources of symmetry breaking.

Main Results:

• Fruit fly behavioral responses demonstrated broken time reversal symmetry.
• Neural network models trained on naturalistic visual data exhibited symmetry breaking, even with symmetric training data.
• Contrast asymmetry and other features of the contrast distribution contribute to symmetry breaking.
• Shallower neural networks showed stronger symmetry breaking than deeper ones.

Conclusions:

• Biological motion detection is not perfectly symmetric upon time reversal.
• Symmetry breaking likely arises from constrained optimization in natural environments.
• Neural network architecture influences the degree of time reversal symmetry breaking.