scholarly journals Continual Competitive Memory

2021 ◽  
Author(s):  
Alexander G. Ororbia
Keyword(s):  
Unsupervised Learning ◽  
Neural System ◽  
Adaptive Resonance Theory ◽  
Competitive Learning ◽  
Sparse Representations ◽  
Classification Problems ◽  
Neural Models ◽  
Resonance Theory ◽  
Adaptive Resonance ◽  
Continual Learning

In this article, we propose a novel form of unsupervised learning that we call continual competitive memory (CCM) as well as a simple framework to unify related neural models that operate under the principles of competition. The resulting neural system, which takes inspiration from adaptive resonance theory, is shown to offer a rather simple yet effective approach for combating catastrophic forgetting in continual classification problems. We compare our approach to several other forms of competitive learning and find that: 1) competitive learning, in general, offers a promising pathway towards acquiring sparse representations that reduce neural cross-talk, and, 2) our proposed variant, the CCM, which is designed with task streams in mind, is needed to prevent the overriding of old information. CCM yields promising results on continual learning benchmarks including Split MNIST and Split NotMNIST.

Bring ART into the ACT

2003 ◽  
Vol 26 (5) ◽  
pp. 610-611 ◽  
Author(s):  
Stephen Grossberg
Keyword(s):  
Cognitive Science ◽  
Real Time ◽  
Adaptive Resonance Theory ◽  
Connectionist Model ◽  
Neural Models ◽  
Connectionist Models ◽  
Resonance Theory ◽  
Adaptive Resonance

ACT is compared with a particular type of connectionist model that cannot handle symbols and use nonbiological operations which do not learn in real time. This focus continues an unfortunate trend of straw man debates in cognitive science. Adaptive Resonance Theory, or ART-neural models of cognition can handle both symbols and subsymbolic representations, and meet the Newell criteria at least as well as connectionist models.

Gearbox Fault Diagnosis Based on Selective Integrated Soft Competitive Learning Fuzzy Adaptive Resonance Theory

2018 ◽  
Vol 19 (1) ◽  
Author(s):  
Xiao-Jin Wan ◽  
Licheng Liu ◽  
Zengbing Xu ◽  
Zhigang Xu
Keyword(s):  
Fault Diagnosis ◽  
Lateral Inhibition ◽  
Adaptive Resonance Theory ◽  
Competitive Learning ◽  
Resonance Theory ◽  
Adaptive Resonance ◽  
Fuzzy Art ◽  
Diagnosis Model ◽  
Fuzzy Adaptive ◽  
Inhibition Theory

In this work, a soft competitive learning fuzzy adaptive resonance theory (SFART) diagnosis model based on multifeature domain selection for the single symptom domain and the single-target model is proposed. In order to solve the problem that the performance of traditional fuzzy ART (FART) is affected by the order of sample input, the similarity criterion of YU norm is introduced into the fuzzy ART network. In the meanwhile, the lateral inhibition theory is introduced to solve the wasteful problem of fuzzy ART mode node. By combining YU norm and lateral inhibition theory with fuzzy ART network, a soft competitive learning ART neural network diagnosis model that allows multiple mode nodes to learn simultaneously is designed. The feature parameters are extracted from the perspectives of time domain, frequency domain, time series model, wavelet analysis, and wavelet packet energy spectrum analysis, respectively. To further improve the diagnostic accuracy, the selective weighted majority voting method is integrated into the diagnosis model. Finally, the selected feature parameters are inputted to the integrated model to complete the fault classification and diagnosis. Finally, the proposed method is verified with a gearbox fault diagnosis test.

Tree-ART2 Learning Model for Spatial Clustering in Second Dimension

2014 ◽  
Vol 543-547 ◽  
pp. 1934-1938
Author(s):  
Ming Xiao
Keyword(s):  
Network Model ◽  
Spatial Data ◽  
Data Clustering ◽  
Clustering Algorithm ◽  
Spatial Clustering ◽  
Adaptive Resonance Theory ◽  
Spatial Distance ◽  
Resonance Theory ◽  
Adaptive Resonance ◽  
Vector Module

For a clustering algorithm in two-dimension spatial data, the Adaptive Resonance Theory exists not only the shortcomings of pattern drift and vector module of information missing, but also difficultly adapts to spatial data clustering which is irregular distribution. A Tree-ART2 network model was proposed based on the above situation. It retains the memory of old model which maintains the constraint of spatial distance by learning and adjusting LTM pattern and amplitude information of vector. Meanwhile, introducing tree structure to the model can reduce the subjective requirement of vigilance parameter and decrease the occurrence of pattern mixing. It is showed that TART2 network has higher plasticity and adaptability through compared experiments.

WIS-ART: UNSUPERVISED CLUSTERING WITH RAM DISCRIMINATORS

1992 ◽  
Vol 03 (01) ◽  
pp. 57-63 ◽  
Author(s):  
Eamon P. Fulcher
Keyword(s):  
Random Access ◽  
Test Phase ◽  
Adaptive Resonance Theory ◽  
The Self ◽  
Hierarchical Level ◽  
Resonance Theory ◽  
Adaptive Resonance ◽  
Pattern Clustering ◽  
Adaptive Pattern ◽  
System Operating

WIS-ART merges the self-organising properties of Adaptive Resonance Theory (ART) with the operation of WISARD, an adaptive pattern recognition machine which uses discriminators of conventional Random Access Memories (RAMs). The result is an unsupervised pattern clustering system operating at near real-time that implements the leader algorithm. ART’s clustering is highly dependent upon the value of a “vigilance” parameter, which is set prior to training. However, for WIS-ART hierarchical clustering is performed automatically by the partitioning of discriminators into “multi-vigilance modules”. Thus, clustering may be controlled during the test phase according to the degree of discrimination (hierarchical level) required. Methods for improving the clustering characteristics of WIS-ART whilst still retaining stability are discussed.

Sign in / Sign up

Export Citation Format

Share Document