Interactive reservoir computing for chunking information streams

Abstract

Chunking is the process by which frequently repeated segments of temporal inputs are concatenated into single units that are easy to process. Such a process is fundamental to time-series analysis in biological and artificial information processing systems. The brain efficiently acquires chunks from various information streams in an unsupervised manner; however, the underlying mechanisms of this process remain elusive. A widely-adopted statistical method for chunking consists of predicting frequently repeated contiguous elements in an input sequence based on unequal transition probabilities over sequence elements. However, recent experimental findings suggest that the brain is unlikely to adopt this method, as human subjects can chunk sequences with uniform transition probabilities. In this study, we propose a novel conceptual framework to overcome this limitation. In this process, neural networks learn to predict dynamical response patterns to sequence input rather than to directly learn transition patterns. Using a mutually supervising pair of reservoir computing modules, we demonstrate how this mechanism works in chunking sequences of letters or visual images with variable regularity and complexity. In addition, we demonstrate that background noise plays a crucial role in correctly learning chunks in this model. In particular, the model can successfully chunk sequences that conventional statistical approaches fail to chunk due to uniform transition probabilities. In addition, the neural responses of the model exhibit an interesting similarity to those of the basal ganglia observed after motor habit formation.

Somatodendritic consistency check for temporal feature segmentation

“…We previously used paired reservoir computing for chunking, where two recurrent networks supervise each other to mimic the partner's responses to a common temporal input 55 . Although that model also learns self-consistency between input data and output data, performance was severely limited since the model required exactly the same number of output neurons as chunks.…”

Fostering Event Compression Using Gated Surprise

“…This model was later extended with an RL agent that controls the gating mechanism with a learned policy [15]. Successfully segmenting the information stream into understandable units was also attempted with reservoir computing [1]. It was shown that this mechanism can be sufficient to identify event boundaries.…”

Create Efficient and Complex Reservoir Computing Architectures with ReservoirPy

“…In addition to setup regular feedback connections as defined in RC, feedback connections mechanism can be hijacked to build teacher Nodes or reward Nodes, and help building architectures based on online learning such as the model from [2] (see sec:asabuki), or help implementing reward-modulated online learning rule like the 3-factor Hebbian learning rule proposed by [11] (ongoing work). These reward or teaching Nodes can be used to provide a connected Node with some target values for training at runtime, even if these targets values are not available before runtime, e.g.…”

“…(see [23] for a recent review): they generally include modified learning methods and architectures and some of them allow for the composition of several reservoirs (decoupled-ESNs [31], tree ESNs [7], deep reservoirs [8], hierarchical-task reservoirs [18], and more exotic architectures like Reservoir-of-Reservoirs (RoR) [4] or self-supervised pairs of reservoirs [2]).…”