11 Encoding and Decoding with Stochastic Neuron models

11.4 Summary

Generalized integrate-and-fire models can predict the spiking activity of cortical cells such as the main inhibitory and excitatory cell type in layer 2-3 of the cortex. For excitatory neurons, more than 80 percent of spike timings can be predicted by these models, while for inhibitory neurons the percentage is close to 100 percent. Similar model performance is seen in the retina, where the activity of up to 250 neurons can be predicted simultaneously (536).

The same models can also be used to decode the activity of neurons. For instance, the spike trains of retinal neurons can be decoded so as to reconstruct a slightly blurred version of the original image movie shown to the retina. Also, the activity of motor cortical neurons can be decoded to reconstruct the intended hand movement in two (or more) dimensions. Thus, the abstract mathematical framework of generalized integrate-and-fire models might ultimately contribute to technical solutions that help human patients.