Adaptive Resonance Theory

Abstract

Adaptive Resonance Theory Definition Adaptive Resonance Theory, or ART, is both a cognitive and neural theory of how the brain quickly learns to categorize, recognize, and predict objects and events in a changing world, and a set of algorithms which computationally embody ART principles and are used in large-scale engineering and technological applications where fast, stable, incremental, learning about complex changing environments is needed. ART clarifies the brain processes from which conscious experiences emerge. It predicts a functional link between processes of Consciousness, Learning, Expectation, Attention, Resonance, and Synchrony (CLEARS), including the prediction that "all conscious states are resonant states." This connection clarifies how brain dynamics enable a behaving individual to autonomously adapt in real time to a rapidly changing world. ART predicts how top-down attention works and regulates fast stable learning of recognition categories. In particular, ART articulates a critical role for "resonant" states in driving fast stable learning; hence the name adaptive resonance. These resonant states are bound together, using top-down attentive feedback in the form of learned expectations, into coherent representations of the world. ART hereby clarifies one important sense in which the brain carries out predictive computation. ART has explained and successfully predicted a wide range of behavioral and neurobiological data, including data about human cognition and the dynamics of spiking laminar cortical networks. ART algorithms have been used in large-scale applications such as medical data base prediction, remote sensing, airplane design, and the control of autonomous adaptive robots.

How Do Expectations Shape Perception?

“…There are several ways in which Bayesian inference can be implemented in a neural architecture [105]. Many frameworks posit that feedback from higher-order areas provides contextual priors [4,79,[106][107][108]. While it may seem natural to implement priors with a feedback process, it should be noted that Bayesian inference can be performed in a purely feedforward manner [109,110].…”

“…Different versions of predictive coding have been developed [117], differing primarily in how the error is computed (by subtraction or division) and how prediction and prediction error neurons are connected [118]. At the same time, there are several other computational theories of perceptual inference that share the computational goal of optimal inference under uncertainty, such as pattern theory [119], adaptive resonance theory [106], particle filtering [79], free energy and active inference [120], and sampling-based probabilistic inference [121,122]. The general motif of these theories is the notion that perceptual inference involves a top-down generative component that predicts and constrains the processing of bottom-up input over time [101].…”

In spoken word recognition the future predicts the past

“…It is also unclear whether this modification would interfere with TRACE's success in accounting for a range of observations in spoken word recognition (see (Gaskell, 2007) for a review). One model proposed to deal with TRACE's shortcoming is Adaptive Resonance Theory (ART): each speech sound produces a resonance wave that is influenced by top-down information until it reaches equilibrium and surfaces to consciousness (Grossberg, 2003). While this theory is consistent with the idea that there is a critical time-limit to receive top-down information, it suggests that there is a linear decay in subphonemic information as temporal distance from the phoneme increases.…”

Information Assurance in Sensor Networks

“…This is a very intuitive approach. The major disadvantage of this approach is that it loses all previous knowledge, therefore suffering "catastrophic forgetting" [47]. In addition, the requirement for storage of all accumulated data sets may not be feasible in many real-world applications due to limited memory and computational resources.…”