Learning a synaptic learning rule

Bengio, Yoshua; Bengio, Samy; Cloutier, Jocelyn

doi:10.1109/ijcnn.1991.155621

Cited by 314 publications

(299 citation statements)

References 0 publications

Supporting

Mentioning

298

Contrasting

Order By: Relevance

“…Using gradient descent, genetic algorithms and simulated annealing, we had already found learning rules for classical conditioning problems, classification problems, and boolean problems (see [7]). Moreover, the experimental results described in section 4 qualitatively agree with learning theory applied to parametric learning rules.…”

Section: Resultsmentioning

confidence: 98%

“…could be defined by equation (2) and | could be a specific vector of rule parameters. If z represents the AND task, then g(z'J(.…”

Section: Extension To Parametric Learning Rulementioning

confidence: 99%

“…Independently of Chalmers, we proposed a similar approach in [2] that generalizes this idea to networks with hidden layers (able to solve non-linearly separable problems) and with the possibility to use any standard optimization methods (such as genetic algorithms, but also gradient descent and simulated annealing). In this paper, we give theoretical principles for the design of such parametric learning rules.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

On the search for new learning rules for ANNs

Bengio

Cloutier

1995

Neural Process Lett

View full text Add to dashboard Cite

show abstract

Section: Resultsmentioning

confidence: 98%

“…could be defined by equation (2) and | could be a specific vector of rule parameters. If z represents the AND task, then g(z'J(.…”

Section: Extension To Parametric Learning Rulementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

On the search for new learning rules for ANNs

Bengio

Cloutier

1995

Neural Process Lett

View full text Add to dashboard Cite

show abstract

“…However, constraints set on learning rules could prevent some from being evolved such as those which include thirdor fourth-order terms. Similar experiments on the evolution of learning rules were also carried out by others [265], [266], [267], [269], [270]. Fontanari and Meir [267] used Chalmers' approach to evolve learning rules for binary perceptrons.…”

Section: B the Evolution Of Learning Rulesmentioning

confidence: 85%

“…Bengio et al's approach [265], [266] is slightly different from Chalmers' in the sense that gradient descent algorithms and simulated annealing, rather than EA's, were used to find near-optimal 's. In their experiments, four local variables and one zeroth-order, three first-order, and three second-order terms in (4) were used.…”

Section: B the Evolution Of Learning Rulesmentioning

confidence: 99%

Evolving artificial neural networks

Yao¹

1999

Proc. IEEE

2,205

100

View full text Add to dashboard Cite

Event‐triggered model reference adaptive control system design for SISO plants using meta‐learning‐based physics‐informed neural networks without labeled data and transfer learning

Duanyai,

Song,

Konghuayrob

et al. 2024

Adaptive Control & Signal

View full text Add to dashboard Cite

SummaryThis paper examines the controllability of a novel Lyapunov‐based model reference adaptive control (MRAC) system designed with a meta‐learning‐based physics‐informed neural network (MLPINN) for linear and nonlinear single‐input and single‐output (SISO) plants without labeled data (MLPINN‐MRAC system). It is devised with the benefits of several techniques: the integration of the identification process in a training mode into an online control mode (straightforward design); no labeled data generation for online identification by a physics‐informed neural network; the prevention of degradation in tracking performance by meta‐learning, as the system triggers a meta‐learning process only when an error threshold detects the deterioration (high efficiency and the reduction of computation cost); quick adaptation to new inputs and an updated control input for each sub‐time span by transfer learning. It is worth noting that the frequency of the meta‐learning event detection significantly affects the step response stability of the nonlinear plant. To achieve a better quality of the stabilization, more frequent event detection is necessary for both beginning and end intervals in control. Sixteen triggering events are enough to shape the acceptable step response of the nonlinear plant, and 44 triggering events achieve the plant's desired step response with its minor modeling error. While, as for the linear plant, a single triggering event is sufficient to attain its tolerable step response and modeling error, implying that intensive event detection is not critical in the identification. It is obvious that the MLPINN‐MRAC system functions well and is more beneficial and efficient for the nonlinear plant.

show abstract

Learning a synaptic learning rule

Cited by 314 publications

References 0 publications

On the search for new learning rules for ANNs

On the search for new learning rules for ANNs

Evolving artificial neural networks

Event‐triggered model reference adaptive control system design for SISO plants using meta‐learning‐based physics‐informed neural networks without labeled data and transfer learning

Contact Info

Product

Resources

About