Global convergence of feedforward networks of learning automata

Phansalkar, V.V.; Thathachar, M. A. L.

doi:10.1109/ijcnn.1992.227089

Cited by 2 publications

(3 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, instead of fixing the hyperplanes of the first level nodes by the initially learned decision tree, as is done in the algorithm in Section IV, we could simultaneously learn the parameters of the hyperplanes in the first layer as well as the Boolean vectors that constitute the weights between the first and second layer. It is possible to use some reinforcement learning algorithms based on LA models for learning hyperplanes [13]. From our prelimenary empirical investigations, it appears that this is a difficult problem.…”

Section: Discussionmentioning

confidence: 99%

“…In what follows, we briefly discuss two stochastic algorithms for learning these Boolean vectors. The first algorithm is based on learning automata (LA) models [13], and the second is based on genetic algorithm (GA) models [14].…”

Section: Decision-tree Pruning As a Boolean-vector Learning Problemmentioning

confidence: 99%

“…The parameter controls the stepsize used in the updating, and it should be sufficiently small. This algorithm is a special case of a general LA algorithm [13]. The general algorithm is known to converge to parameter values that are local maxima of expected value of calculated in step 7 of the algorithm.…”

Section: A La Pruningmentioning

confidence: 99%

See 2 more Smart Citations

New algorithms for learning and pruning oblique decision trees

Shah

Sastry

1999

IEEE Trans. Syst., Man, Cybern. C

View full text Add to dashboard Cite

Abstract-In this paper, we present methods for learning and pruning oblique decision trees. We propose a new function for evaluating different split rules at each node while growing the decision tree. Unlike the other evaluation functions currently used in literature (which are all based on some notion of purity of a node), this new evaluation function is based on the concept of degree of linear separability. We adopt a correlation-based optimization technique called the Alopex algorithm for finding the split rule that optimizes our evaluation function at each node. The algorithm we present here is applicable only for 2-class problems. Through empirical studies, we demonstrate that our algorithm learns good compact-decision trees.We suggest a representation scheme for oblique decision trees that makes explicit the fact that an oblique decision tree represents each class as a union of convex sets bounded by hyperplanes in the feature space. Using this representation, we present a new pruning technique. Unlike other pruning techniques, which generally replace heuristically selected subtrees of the original tree by leaves, our method can radically restructure the decision tree. Through empirical investigation, we demonstrate the effectiveness of our method.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Decision-tree Pruning As a Boolean-vector Learning Problemmentioning

confidence: 99%