Unsupervised context-based learning of multiple temporal sequences

A self-organizing neural network is proposed to handle multiple temporal sequences with states in common. The proposed network combines context-based competitive learning with time-delayed Hebbian learning to encode spatial features and temporal order of sequence items, respectively. A responsibility function to avoid catastrophic forgetting, and a redundancy mechanism to avoid loss of stored sequences increase the reliability of the model. States shared by different sequences are encoded into a single neuron, whereas, for recall, context information indicates the correct sequence to be followed in the case of ambiguity. Simulations with trajectories of a PUMA 560 robot are performed to test the network accuracy, robustness to noise and tolerance to faults.