| Principles of real-time computing with feedback applied to cortical microcircuit models (2006) | |||||||||||||||||
Abstract | |||||||||||||||||
| What happens if one allows feedback from trained readouts? Figure 1: Expanding the “liquid computing model ” of [3] by allowing feedback from trained linear readouts. The circuit itself is a generic recurrent circuit of spiking neurons, based on biological data (not constructed for any particular task). Or equivalently: What happens if one trains a neuron within a generic cortical microcircuit model? Figure 2: Modifying the weights of synaptic connections to a readout neuron that provides feedback (left panel) is essentially equivalent to modifying the weights to a single neuron within the otherwise gereric cortical microcircuit model (right panel). It had previously been shown that generic cortical microcircuit models can perform complex real-time computations on continuous input streams, provided that these computations can be carried out with a rapidly fading memory. We investigate in this article the computational capability of such circuits in the more realistic case where not only readout neurons, but | |||||||||||||||||
Publication details | |||||||||||||||||
| |||||||||||||||||