Learning algorithms for oscillatory novelist circuits

摘要

The learning algorithm for the oscillatory linear oscillator circuit is described. In one embodiment, the neuromodule circuit learning network comprises a plurality of neuromodule circuit nodes, each comprising a recognized neuron unit and a generated neuron unit. The learning network may include a plurality of neuromodule circuit feedforward couplings between the recognition neuron units within the neuromodic circuit nodes and a plurality of neodromic circuit feedbacks between the generated neuron units in the neuromodic circuit nodes Coupling. / RTI The learning network may also include learning configured to train activity among the recognized neuron units, train the generated neuron unit for learning in one mode, induce activity among the generated neuron units, and train the recognized neuron unit for learning in another mode Controller. Various in-depth learning algorithms can be implemented in the learning network. Two examples include a non-supervised New Lomopic circuit network and a wake-slip algorithm for target propagation.