Monday 02 Oct 2017: Dynamics Seminar: The whence and whither of a simplified model of an auditory neuron
Dan Goodman - Imperial College London
LSI Seminar Room B 14:30-15:30
There is an ongoing debate in neuroscience about the appropriate level of detail of neural models. Increasingly, there has been a trend towards extremely complex and detailed models. In this talk, I'll argue against this trend: simpler models are easier to interpret and contribute more to the development of neuroscience. I'll illustrate this with an example from my recent work creating an incredibly simplified model of one type of auditory neuron (chopper cells in the ventral cochlear nucleus). I'll explain how this simple model arises in a principled way. The key simplifying assumption I'll use is that in this study I was only interested in the ongoing, stationary response of the neuron, not its dynamics at the onset of the stimulus. This assumption allowed a huge reduction in the complexity of the model. I'll then show that despite its simplicity, the model was able to provide a new view on an old debate on the neural mechanisms underlying the behaviour of chopper cells, as well as explaining an existing data set that had not previously been explained, and making robust predictions for the effect of the loss of input synapses that has been hypothesised to be a component in certain types of hearing loss.