English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Attractor dynamics makes the time evolution of a dynamical system collapse into a subspace of its available state space; in the simplest instance, the one which will concern us, the subspace is a set of discrete points (or trajectories constrained to their neighborhood). To the extent that the attractor endpoint depends on the initial conditions of the dynamics, the state space is partitioned into basins for each attractor state. In the cognitive metaphor, a class of stimuli (initial states falling into the same basin) is defined by the fact that all its members cause the retrieval of the same prototypical representation (the attractor state). The stimulus and the retrieved representation belong to the same state space; for the recurrent neural networks that we will consider, each point in the state space is a distribution of activity in the network: the one initially imposed by a stimulus evolves into a selective self-sustained activity pattern (i.e. one that depends on the initial state, but not on its details). Attractor neural models have been initially developed and improved to account for a wide array of experimental evidence related to working memory. It is increasingly becoming clear that the dynamic scheme has a wider scope. Models based on bistable (two attractors) or multi-stable (multiple attractors) networks have been proposed as theoretical underpinnings for understanding perceptual decision mechanisms and processes of information integration, as well as multi-stable perception and binocular rivalry. Motivated by such a multiplicity of putative roles that attractor dynamics might play in neural computation, and the related interest in versatile components for computation in neuromorphic systems, we undertook to develop VLSI neural networks, communication infrastructures, and strategies for experiments and analysis, suited to match this level of complexity. The core of our systems is the neuromorphic chip described in detail in [8] and [9], comprising 128 Integrate-and-fire (IF) neurons and about 16384 plastic reconfigurable synapses.

A distributed VLSI attractor network learning visual stimuli in real time and performing perceptual decisions