Gallant, S.I.: Perceptron-Based Learning Algorithms, in IEEE Transactions on Neural Networks, 1 (1990) 179-191  Home/Search   Document Not in Database   Summary   Related Articles   Check

....structures we are interested in are feedforward networks with cascading hidden layers. These are unique as network units are a function of both the input space and other hidden units. Algorithms producing cascading architectures, such as the cascade correlation algorithm [2] the Tower algorithm [3] and variants [5] sequentially tenure units to minimise the classification, or mean square error over a training set. While all of these algorithms are effective, the cascade correlation algorithm has a distinct disadvantage for decompositional rule extraction. A cascade correlation network ....

....sufficiently approximates the logical function we are looking for and the network is complete; or the unit is added to the set with its weights frozen, and the process is repeated by connecting and training a new unit. The Tower algorithm trains each unit using the Pocket al..gorithm with ratchet [3]. This induces two half spaces over the input space biased to classify as many positive examples as possible. Bptower, rather than using Pocket al..gorithm, uses gradient descent to induce the half spaces. The termination condition for Bptower depends on the root mean square error found in ....

S. Gallant, `Perceptron-based learning algorithms', IEEE Transactions on Neural Networks, 1(2), 179--191, (June 1990).

....by [33] During constructive learning, single neurons and links, or whole layers or subnetworks, can be added or deleted from the network architecture. Most well known are probably the algorithms which incrementally add hidden units to a neural network, such as Cascade Correlation [9] or Tower [13] algorithms. A large amount of work has been done to investigate these types of algorithms, see for example [28, 32] The study of [6] investigated combinations of Cascade Correlation or Tower algorithms with rule extraction from neural network methods, and clarified some issues on combining rule ....

S. Gallant. Perceptron-based learning algorithms. IEEE Transactions on Neural Networks, 1(2):179--191, 1990.

....threshold neurons for pattern classification. 1) Constructive Learning Using Iterative Weight Update: A number of algorithms that incrementally construct networks of threshold neurons for learning the binary to binary mapping have been proposed in the literature (for example, the tower, pyramid [19], tiling [31] upstart [15] oil spot [29] and sequential [28] algorithms) These algorithms differ in terms of their choices regarding: restrictions on input representation (e.g. binary or bipolar valued inputs) when to add a neuron; where to add a neuron; connectivity of the added neuron; ....

....is not linearly separable then the perceptron algorithm behaves poorly in the sense that the classification accuracy on the training set can fluctuate widely from one training epoch to the next. Several modifications to the perceptron algorithm (e.g. the pocket al..gorithm with ratchet modification [19], the thermal perceptron algorithm [16] the loss minimization algorithm [24] and the barycentric correction procedure [38] are proposed to find a reasonably good weight vector that correctly classifies a large fraction of the training set when is not linearly separable and to converge to zero ....

[Article contains additional citation context not shown here]

S. Gallant, "Perceptron based learning algorithms," IEEE Trans. Neural Networks, vol. 1, pp. 179--191, 1990.

....obtained by LocBoost and RealBoost (confidence rated AdaBoost) 20] for three datasets from the UCI repository. The figures are the 5 fold cross validated error results obtain by LocBoost and RealBoost, respectively, where the weak classifier used by RealBoost is the perceptron pocket al..gorithm [10]. The LocBoost algorithm was applied with up to 10 boosting rounds and 5 EM iterations per step. RealBoost was applied with 50 boosting rounds in all runs and in each boosting round the pocket al..gorithm was trained for 1000 iterations. Note that these preliminary results are for illustrative ....

S. Gallant. Perceptron-based learning algorithms. IEEE Trans. Neural Networks, 1(2):179-191, 1990.

....methods in order to find a good genetic network. In the next section, we will propose a practical algorithm, inspired by the HAKKE system [5, 8] We also note that our problem is very closely related to a classical problem of learning by a perceptron, where some learning algorithms are proposed [4, 6, 7]. 145 Algorithm AK algorithm Given S: set of linear inequalities over variables X begin while S #= # do if there are inequalities in S that contains a single variable then U : x # X x occurs as a single variable in at least one inequalities of S ; pick at random y # U ; F (y) ....

Gallant, S. I., Perceptron-based learning algorithms, IEEE Trans. on Neural Networks, 1:179--191, 1990.

....er. The weights associated with the feature set need to be trained to orient the hyper plane of separation in k dimensional space to best classify the samples. This is done using a pocket variant of the perceptron algorithm, which is especially suited for linearly separable classes, suggested in [3]. Unlike the BIOKDD01: Workshop on Data Mining in Bioinformatics (with SIGKDD01 Conference) page 10 original perceptron training algorithm which does not guarantee convergence for classes that are not linearly separable, the pocket al..gorithm always converges to the optimal solution, that is, the ....

.... (with SIGKDD01 Conference) page 10 original perceptron training algorithm which does not guarantee convergence for classes that are not linearly separable, the pocket al..gorithm always converges to the optimal solution, that is, the one that produces the least number of misclassi cations [3]. The unmodi ed perceptron has been tried in [8] without any consideration for prior feature selection, and its performance was found to be disappointing. The excellent results we achieve using the perceptron is a validation of our prioritized feature selection approach. An outline of this ....

S.I.Gallant. Perceptron based learning algorithms. IEEE Transactions on Neural Networks, 1(2):179-191, 1990.

....methods in order to find a good genetic network. In the next section, we will propose a practical algorithm, inspired by the HAKKE system [5, 8] We also note that our problem is very closely related to a classical problem of learning by a perceptron, where some learning algorithms are proposed [4, 6, 7]. 145 Algorithm AK algorithm Given S: set of linear inequalities over variables X begin while S 6= OE do if there are inequalities in S that contains a single variable then U : fx 2 X j x occurs as a single variable in at least one inequalities of Sg ; pick at random y 2 U ; F (y) fe 2 ....

Gallant, S. I., Perceptron-based learning algorithms, IEEE Trans. on Neural Networks, 1:179--191, 1990.

....obtained by LocBoost and RealBoost (confidence rated AdaBoost) 20] for three datasets from the UCI repository. The figures are the 5 fold cross validated error results obtain by LocBoost and RealBoost, respectively, where the weak classifier used by RealBoost is the perceptron pocket al..gorithm [10]. The LocBoost algorithm was applied with up to 10 boosting rounds and 5 EM iterations per step. RealBoost was applied with 50 boosting rounds in all runs and in each boosting round the pocket al..gorithm was trained for 1000 iterations. Note that these preliminary results are for illustrative ....

S. Gallant. Perceptron-based learning algorithms. IEEE Trans. Neural Networks, 1(2):179-191, 1990.

....probabilities. The situation will be the same whenever a binary output representation is taken. Thus, they should be avoided if useful solutions are required. Of course, unary representations are not the only possible choice to find useful solutions. For example, a thermometer representation [5] for the same problem could be 8 : z (1) j (0; 0; 0) z (2) j (1; 0; 0) z (3) j (1; 1; 0) z (4) j (1; 1; 1) 4.4) which has as solution 8 : P (z (1) jx) 1 Gamma o 1 (x) P (z (2) jx) o 1 (x) Gamma o 2 (x) P (z (3) jx) o 2 (x) Gamma ....

S. Gallant, "Perceptron-based learning algorithms," IEEE Trans. Neural Networks 1 , 179 (1990).

No context found.

Gallant, S.I.: Perceptron-Based Learning Algorithms, in IEEE Transactions on Neural Networks, 1 (1990) 179-191

No context found.

S.I. Gallant, "Perceptron-Based Learning Algorithms," in IEEE Transactions on Neural Networks, vol. 1, no. 2, pp. 179-191, June 1990.

No context found.

Gallant, S.I.: Perceptron-Based Learning Algorithms, in IEEE Transactions on Neural Networks, 1 (1990) 179-191

No context found.

S.I. Gallant, "Perceptron-Based Learning Algorithms," in IEEE Transactions on Neural Networks, vol. 1, no. 2, pp. 179-191, June 1990.

No context found.

S.I. Gallant, Perceptron-Based Learning Algorithms , IEEE Transactions on Neural Networks, 1 (2), pp. 179-191, 1990.

No context found.

Gallant, S.I.: Perceptron-Based Learning Algorithms, in IEEE Transactions on Neural Networks, vol. 1, no. 2 (1990) 179-191

No context found.

S.I. Gallant, "Perceptron-Based Learning Algorithms", IEEE Transactions on Neural Networks, 1 (2), pp. 179-191, 1990.

No context found.

S.I. Gallant, "Perceptron-Based Learning Algorithms," in IEEE Transactions on Neural Networks, vol. 1, no. 2, pp. 179-191, June 1990.

No context found.

S.I. Gallant, PerceptronBased Learning Algorithms. IEEE Transactions on Neural Networks, 1 (1990) 179--191.

No context found.

Gallant, S. I. Perceptron-based learning algorithms. IEEE Transactions on Neural Networks 1 (1990), 179--191.

No context found.

Gallant, S. I. Perceptron-based learning algorithms. IEEE Transactions on Neural Networks 1 (1990), 179--191.

No context found.

S.I. Gallant. Perceptron-Based Learning Algorithms. IEEE Trans. on Neural Networks, 1(2):179 -- 191, 1990.

No context found.

S.I. Gallant. Perceptron-based learning algorithms. IEEE Transactions on Neural Networks, 1(2):179--191, June 1990.

No context found.

Gallant, S. I. Perceptron-based learning algorithms. IEEE Transactions on Neural Networks 1 (1990), 179--191.

No context found.

S. I. Gallant. Perceptron-based learning algorithms. IEEE Trans. on Neural Networks, 1:179--191, 1990.

No context found.

Stephen I. Gallant, "Perceptron-based learning algorithms," IEEE Transactions on Neural Networks, vol. 1, pp. 179--191, 1990.

First 50 documents  Next 50

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