Find in Library

The ability to learn and perform statistical inference with biologically plausible recurrent network of spiking neurons is an important step towards understanding perception and reasoning. <br/>Here we derive and investigate a new learning rule for recurrent spiking networks with hidden neurons, combining principles from variational learning and reinforcement learning. <br/>Our network defines a generative model over <br/>spike train histories and the derived learning rule has the form of a <br/>local Spike Timing Dependent Plasticity rule modulated by global factors (neuromodulators) conveying information about ``novelty on a statistically rigorous ground.<br/>Simulations show that our model is able to learn both<br/>stationary and non-stationary patterns of spike trains.<br/>We also propose one experiment that could potentially be performed with animals in order to test the dynamics of the predicted novelty signal.