Method for unsupervised sequence learning using reinforcement learning and neural networks

A non-transitory computer-readable storage medium storing executable computer program instructions for training a sequence learning model based on reinforcement learning and neural network, the computer program instructions comprising instructions for:

• retrieving input sequence data, the input sequence data including one or more input time sequences;
• encoding the input sequence data into output symbol data using a sequence learning model, the output symbol data including one or more symbolic representations;
• decoding, based on a neural network, the output symbol data to decoded sequence data, the decoded sequence data including one or more decoded time sequences that are to match the one or more input time sequences in the input sequence data;
• comparing the decoded sequence data with the input sequence data;
• and
• updating the sequence learning model based on the comparison.

A computer system for training a sequence learning model based on reinforcement learning and neural network, the system comprising:

• a processor;
• and
• memory storing an encoder for:
• retrieving input sequence data, the input sequence data including one or more input time sequences;
• and
• encoding the input sequence data into output symbol data using a sequence learning model, the output symbol data including one or more symbolic representations;
• a decoder for:
• decoding, based on a neural network, the output symbol data to decoded sequence data, the decoded sequence data including one or more decoded time sequences that are to match the one or more input time sequences in the input sequence data;
• and
• comparing the decoded sequence data with the input sequence data;
• and
• the encoder also for:
• updating the sequence learning model based on the comparison.