Decoupled extended Kalman filter training of feedforward layered networks

Puskorius, G.V.; Feldkamp, L.A.

doi:10.1109/ijcnn.1991.155276

Cited by 170 publications

(102 citation statements)

References 6 publications

Supporting

Mentioning

101

Contrasting

Unclassified

Order By: Relevance

“…This choice differs from that made in [4] and [11], where the constraints imposed by the use of neural network are not particularly taken into account, and where the diagonal terms of P 0 are chosen of the order of 100, r being also set to value close to 1.…”

Section: Proposed Algorithm For the Training Of Feedforward Neural Momentioning

confidence: 98%

“…for which the weights must converge to constant values, they should not be allowed to drift. Nevertheless, a diagonal matrix Q with positive components is sometimes added in the update equation for P(k) both to avoid numerical instability, and to help to avoid local minima [11] [12]. But the adequate order of magnitude of the elements of Q which simultaneously makes the weights converge to constant values seems to be highly problem-dependent.…”

Section: Remarkmentioning

confidence: 99%

“…and which was also modeled in [11]. The process is described by a deterministic third-order single-input, single-output model:…”

Section: Modeling a Third-order Deterministic Process (Dynamic 5-inpumentioning

confidence: 99%

“…such that the initial output values of the hidden neurons are in the linear part of the tanh. We propose to initialize the weights with small random values, a choice which is also made in most EKF approaches (in [4], [11] and [12] for example).…”

Section: Proposed Algorithm For the Training Of Feedforward Neural Momentioning

confidence: 99%

“…to the estimation of their weights, both for feedforward networks [14] [11], and for recurrent networks [7] [15].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A recursive algorithm based on the extended Kalman filter for the training of feedforward neural models

Rivals¹,

Personnaz²

1998

Neurocomputing

View full text Add to dashboard Cite

To cite this version:Isabelle Rivals, Léon Personnaz. A recursive algorithm based on the extended Kalman filter for the training of feedforward neural models. Neurocomputing, Elsevier, 1998, 20 (1-3), pp.279-294. In Neurocomputing (1-3): 279-294 (1998 AbstractThe Extended Kalman Filter (EKF) is a well known tool for the recursive parameter estimation of static and dynamic nonlinear models. In particular, the EKF has been applied to the estimation of the weights of feedforward and recurrent neural network models, i.e. to their training, and shown to be more efficient than recursive and non recursive first-order training algorithms; nevertheless, these first applications to the training of neural networks did not fully exploit the potentials of the EKF.In this paper, we analyze the specific influence of the EKF parameters for modeling problems, and propose a variant of this algorithm for the training of feedforward neural models which proves to be very efficient as compared to non recursive second-order algorithms. We test the proposed EKF algorithm on several static and dynamic modeling problems, some of them being benchmark problems, and which bring out the properties of the proposed algorithm..

show abstract

Section: Proposed Algorithm For the Training Of Feedforward Neural Momentioning

confidence: 98%

Section: Remarkmentioning

confidence: 99%

“…and which was also modeled in [11]. The process is described by a deterministic third-order single-input, single-output model:…”

Section: Modeling a Third-order Deterministic Process (Dynamic 5-inpumentioning

confidence: 99%

Section: Proposed Algorithm For the Training Of Feedforward Neural Momentioning

confidence: 99%

“…to the estimation of their weights, both for feedforward networks [14] [11], and for recurrent networks [7] [15].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A recursive algorithm based on the extended Kalman filter for the training of feedforward neural models

Rivals¹,

Personnaz²

1998

Neurocomputing

View full text Add to dashboard Cite

show abstract

Kalman Filtering and Neural Networks

Haykin¹

2001

1,225

View full text Add to dashboard Cite

The Unscented Kalman Filter

Wan

Merwe

2001

Kalman Filtering and Neural Networks

822

513

View full text Add to dashboard Cite

Decoupled extended Kalman filter training of feedforward layered networks

Cited by 170 publications

References 6 publications

A recursive algorithm based on the extended Kalman filter for the training of feedforward neural models

A recursive algorithm based on the extended Kalman filter for the training of feedforward neural models

Kalman Filtering and Neural Networks

The Unscented Kalman Filter

Contact Info

Product

Resources

About