Contrastive Hebbian Learning in the Continuous Hopfield Model

Movellan, Javier R.

doi:10.1016/b978-1-4832-1448-1.50007-x

Cited by 89 publications

(98 citation statements)

References 10 publications

(15 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When the diffusion term vanishes, the network becomes deterministic and goodness can only increase through time. Since the goodness function is bounded upward, the activations stabilize at local maxima of G. It is also known that this deterministic kernel (when (z = 0) is trainable with the CHL rule, but that instabilities may occur due to the existence of multiple maxima in the G function (Hinton, 1989;Movellan, 1990;Peterson & Anderson, 1987;Peterson & Hartman, 1989).…”

Section: Activation Dynamicsmentioning

confidence: 99%

“…Adaptive gains may prove important in hardware implementations with limited precision weights but are not particularly relevent for our simulations. As discussed in Movellan (1990), gradient descent calls for the self-connections to be changed at half the rate of the other weights. Our simulations followed this rule.…”

Section: Simul~ionsmentioning

confidence: 99%

“…In a previous article (Movellan and McClelland, in press), we presented initial work with the CHL algorithm and another instantiation of the principles of continuous, stochastic, interactive processing. Our approach there was based on the ideas of contrastive learning (Baldi & Pineda, 1991;Movellan, 1990) rather than Markovian diffusion theory. It should be noted that SBM can learn discrete binary probability distributions.…”

Section: Introductionmentioning

confidence: 99%

“…CHL is a general learning rule previously applied to a variety of models including the discrete Hopfield model (Hopfield, Feinstein, 8c Palmer, 1983), the original stochastic Boltzmann machine (Ackley et al, 1985), the harmonimum (Smolensky, 1986), the deterministic Boltzmarm machine (Galland & Hinton, 1989;Hinton, 1989;Peterson & Anderson, 1987), and the continuous Hopfield model (Movellan, 1990). Here we show that in SDNs, the CHL rule performs gradient descent on an error function that captures differences between entire distributions.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Untitled

1993

Cognitive Science

View full text Add to dashboard Cite

Section: Activation Dynamicsmentioning

confidence: 99%

Section: Simul~ionsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Untitled

1993

Cognitive Science

View full text Add to dashboard Cite

“…The network was trained to be able to associate an individual's face, semantics, and name whenever one of these was presented, using the Contrastive Hebbian Learning (CHL) algorithm (Movellan, 1990).…”

mentioning

confidence: 99%

Dissociated Overt and Covert Recognition as an Emergent Property of Lesioned Attractor Networks

Fraah¹,

O’Reilly²,

Vecera³

1992

View full text Add to dashboard Cite

Learning Stiffness Tensors in Self‐Activated Solids via a Local Rule

Tang,

Ye,

Jia

et al. 2024

Advanced Science

View full text Add to dashboard Cite

Mechanical metamaterials are often designed with particular properties for specific load‐bearing functions. Alternatively, this study aims to create a class of active lattice metamaterials, dubbed self‐activated solids, that can learn desired stiffness tensors from the elastic deformations they experienced, a crucial feature to improve the performance, efficiency, and functionality of materials. Artificial adaptive matters that combine sensory, control, and actuation elements can offer appealing solutions. However, challenges still remain: The designs will rely on accurate off‐line and global computations, as well as intricate coordination among individual elements. Here, a simple online and local learning strategy is initiated based on contrastive Hebbian learning to gradually guide self‐activated solids to possess sought‐after stiffness tensors autonomously and reversibly. During learning, the bond stiffness of the active lattice varies depending only on its local strain. The numerical tests show that the self‐activated solid can not only achieve the desired bulk, shear, and coupling moduli but also manifest uni‐mode and bi‐mode extremal materials by itself after experiencing the corresponding elastic deformations. Further, the self‐activated solid can also achieve the desired time‐varying moduli when exposed to temporally different loads. The design is applicable to any lattice geometries and is resistant to damage and instabilities. The material design approach and the physical learning strategy suggested can benefit the design of autonomous materials, physical learning machines, and adaptive robots.

show abstract

Contrastive Hebbian Learning in the Continuous Hopfield Model

Cited by 89 publications

References 10 publications

Untitled

Untitled

Dissociated Overt and Covert Recognition as an Emergent Property of Lesioned Attractor Networks

Learning Stiffness Tensors in Self‐Activated Solids via a Local Rule

Contact Info

Product

Resources

About