Learning Sequence Attractors in Recurrent Networks with Hidden Neurons
This work addresses the challenge of understanding sequence memory and temporal information processing in the brain, which is an incremental advance in computational neuroscience.
The study tackled the problem of how recurrent networks of binary neurons can learn to store and retrieve predefined pattern sequences, showing that hidden neurons are necessary for storing arbitrary sequences and developing a local learning algorithm that converges to sequence attractors, with robust performance demonstrated on synthetic and real-world datasets.
The brain is targeted for processing temporal sequence information. It remains largely unclear how the brain learns to store and retrieve sequence memories. Here, we study how recurrent networks of binary neurons learn sequence attractors to store predefined pattern sequences and retrieve them robustly. We show that to store arbitrary pattern sequences, it is necessary for the network to include hidden neurons even though their role in displaying sequence memories is indirect. We develop a local learning algorithm to learn sequence attractors in the networks with hidden neurons. The algorithm is proven to converge and lead to sequence attractors. We demonstrate that the network model can store and retrieve sequences robustly on synthetic and real-world datasets. We hope that this study provides new insights in understanding sequence memory and temporal information processing in the brain.