M. Riedmiller and H. Braun. RPROP: A fast and robust backpropagation learning strategy. In Marwan Jabri, editor, Fourth Australian Conference on Neural Networks, pages 169 - 172, Melbourne, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....A predictor can be implemented by an Artificial Neural Network (ANN) which uses a certain number of past samples to predict a future event trying to track the signal. To implement the tracking ANN, a multilayer perceptron architecture was chosen. The ANN was trained with Resilient Propagation [10] with N input, H hidden nodes and 1 output (N Gamma H Gamma 1) The symbols to be encoded are given by: e n = xn Gamma g(xn Gamma1 ; xn GammaN ) 20) where g(xn Gamma1 ; xn GammaN ) represents the output of the ANN. The ANN was trained with a set of patterns comprising N past ....

.... 58.1 (58.3) 57.8) Filter 58.9 ANN 56.0 56.1 55.8 ANN 56.7 56.7 56.6 LPC 60.9 (61.0) 60.6) LPC 61.4 Adapt LMS 58.1 (58.3) 57.9) Adapt LMS 58.8 Adapt SGN 58.5 (58.6) 58.4) Adapt SGN 59.2 perceptron with architecture 4 2 1, and was trained with Resilient Propagation [10] with a learning rate equal to 0.1. The inputs to the network are the 4 past values of the signal (xn Gamma1 ; xn Gamma2 ; xn Gamma3 ; xn Gamma4 ) while what the network is trained to produce is the actual value (x n ) As the EEGs available for training provide a huge number of patterns, they ....

H.Braun, M. Riedmiller, "Rprop: A fast and robust Backpropagation Learning Strategy", Proc. ACNN, 1993.

....be evaluated. So, the main idea of this work is to apply a sophisticated optimization technique for parameter identification which enables the use of the heuristic search even when being applied to high dimensional problems. For this purpose, we chose resilient propagation (RPROP) Riedmiller and Braun, 1993; Braun and Riedmiller, 1993; Zell et al. 1994) because it is easy to apply (only needs first derivatives) but has a performance like second order methods as for example the Levenberg Marquardt algorithm (Hagan and Menhaj, 1994) The following sections provide a description of the fuzzy model, ....

....So, the main idea of this work is to apply a sophisticated optimization technique for parameter identification which enables the use of the heuristic search even when being applied to high dimensional problems. For this purpose, we chose resilient propagation (RPROP) Riedmiller and Braun, 1993; Braun and Riedmiller, 1993; Zell et al. 1994) because it is easy to apply (only needs first derivatives) but has a performance like second order methods as for example the Levenberg Marquardt algorithm (Hagan and Menhaj, 1994) The following sections provide a description of the fuzzy model, the identification procedure ....

[Article contains additional citation context not shown here]

Braun, H. and M. Riedmiller (1993). Rprop: A fast and robust backpropagation learning strategy. In: Proceedings of the ACNN.

....Our current approach uses local statistical features as average and variance plus stereo disparity and motion patterns, in the vicinity of sampled positions. Identification is done using an associative memory implemented by a multi layer perceptron with a back propagation training algorithm [1] (BPNN) The BPNN maps the features to an address in the long term memory (LTM) which stores various other information. If the representation is new, supervised learning is automatically invoked, inserting the new feature set into LTM and updating retraining the BPNN. Once a representation is ....

H. BRAUN and M. RIEDMILLER. Rprop: A fast and robust backpropagation learning strategy. In Proc. of the International Conference on Neural Networks, pages 123--134, 1993.

....After this reduction we compared the results of different neural network algorithms to achieve best results. In detail, we tried Radial Basis Functions (RBF) Pog89] Counterpropagation [HN88] and Feedforward networks like Backpropagation (BackProp) Rum86] and Resilient Propagation (RProp) [Bra93]. For all networks, we first trained with half of the data samples (2430) and all components (96) and used the rest of the 4860 data samples for the validation set. After that we reduced the components of the input vectors to the number of 10. The main reason was to achieve a good and ....

H. Braun and M. Riedmiller. Rprop: A fast and robust backpropagation learning strategy. In ACNN, pages 598--591, 1993.

.... [HN89] we have used the geometrical interpolation again ( G o93] and [Lud95] The results of this algorithm have been compared with results of counterpropagation (CPN) topological interpolation in the outstar layer of a counterpropagation net [G o92] and resilient propagation (Rprop) Bra93] 3.2 SOM for the analysis of PCM data For the problem of analyzing PCM data we also used the self organizing map. The map had 35 neurons and was rectangular shaped (7x5) All of the 595 data samples have been used to train Figure 3: Topological structure of chips on a wafer map. All chips ....

H. Braun and M. Riedmiller. Rprop: A fast and robust backpropagation learning strategy. In ACNN, pages 598--591, 1993.

....range and prevents problems like floating point overflow errors during computations. Values of are usually 1:75 : 2:25 and typically assumes low values like 0:0001 because QuickProp is very sensitive to it. The final algorithm that we consider is called Resilient backpropagation (RProp)[1] because it uses the local topology of the error surface to make a more appropriate weight change. In other words, we introduce a personal update value for each weight, which evolves during the learning process according to its local view of the error function. Thus we have two sets of learning ....

H. Braun and M. Riedmiller. Rprop : A Fast and Robust Backpropagation Learning Strategy. In Proceedings of the ACNN, 1993.

....back propagation (RPROP) algorithm was used to train the networks used in this dissertation, since it gives relatively fast convergence, yet is easy to implement. This algorithm is a form of error back propagation in which a separate update step size is maintained for each parameter in the network [25, 140]. For each input vector the sensitivity of the network s output error with respect to each of its parameters is determined by computing the partial derivatives of the error with respect to these parameters: E w ij ; E b j (4.12) where the error E is as defined in Equation 4.11, and w ij and ....

H. Brann and M. Riedmiller. "Rprop: A fast and robust backpropagation learning strategy". In ACNN, 1993.

....surface to compute the weight change. QuickProp approximates the error surface to be locally quadratic and attempts to jump in one step from the current position directly into the minimum of the quadratic. ffl RProp The final algorithm that we consider is called Resilient backpropagation (RProp)[52] because it uses the local topology of the error surface to make a more appropriate weight change. In other words, we introduce a personal update value for each weight, which evolves during the learning process according to its local view of the error function. RProp is very powerful and ....

H. Braun and M. Riedmiller, "Rprop : A Fast and Robust Backpropagation Learning Strategy," in Proceedings of the ACNN, 1993.

No context found.

M. Riedmiller and H. Braun. RPROP: A fast and robust backpropagation learning strategy. In Marwan Jabri, editor, Fourth Australian Conference on Neural Networks, pages 169 - 172, Melbourne, 1993.

No context found.

M. Riedmiller and H. Braun. RPROP: A fast and robust backpropagation learning strategy. In Marwan Jabri, editor, Fourth Australian Conference on Neural Networks, pages 169 - 172, Melbourne, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC