A reinforcement learning framework for spiking networks with dynamic synapses
Computational Intelligence and Neuroscience
| Hi-index | 0.00 |
In this research paper, based on biological motivation, synapse is modeled as a Finite Impulse Response (FIR) linear filter. Motivation for the concept of robust neural networks is proposed. A novel incremental gradient based learning algorithm is derived (to update the FIR filter coefficients in successive slots). This model of neuron is utilized for arriving at a multi-layer perceptron. Also, a potential associative memory based on FIR filter model of synapse is proposed. Briefly the novel model of neuron is compared with the traditional model of neuron. It is reasoned that the traditional model of neuron is very restrictive.