Temporal Pattern Recognition in the Retina

Abstract

The brain processes stimuli from the environment and seeks patterns in order to make predictions about the future. The biophysical mechanism for pattern recognition in the nervous system is an ongoing problem, and the retina serves as a tractable model system where we can begin to answer these questions, for it is isolated from the rest of the brain and provides for easy experimental access. Following a periodic sequence of ashes, certain ganglion cells have been shown to fire a burst of spikes following the omission or breaking of a pattern rather than when the pattern kept going as expected, suggesting that the retina recognizes the pattern and conveys to the brain when the pattern has stopped its periodicity, thereby conserving energy. This thesis investigates this "Omitted Stimulus Response" (OSR)to sequences of variable ashes drawn from a probability distribution and observed certain retinal ganglion cells which re after a randomly generated repeating periodic sequence is broken. The proposed model for OSR generation is quantitatively presented, along with new unpublished data which suggests it is wrong. New models based on the generalized linear model and a multiplicative model based on synaptic plasticity are developed and tested, and the multiplicative model is shown to accurately predict the OSR response of retinal ganglion cells.