Extracting M-of-N rules from trained neural networks Export

@article{Setiono:2000, abstract = {An effective algorithm for extracting M-of-N rules from trained feedforward neural networks is proposed. First, we train a network where each input of the data can only have one of the two possible values, -1 or one. Next, we apply the hyperbolic tangent function to each connection from the input layer to the hidden layer of the network. By applying this squashing function, the activation values at the hidden units are effectively computed as the hyperbolic tangent (or the sigmoid) of the weighted inputs, where the weights have magnitudes that are equal one. By restricting the inputs and the weights to binary values either -1 or one, the extraction of M-of-N rules from the networks becomes trivial. We demonstrate the effectiveness of the proposed algorithm on several widely tested datasets. For datasets consisting of thousands of patterns with many attributes, the rules extracted by the algorithm are simple and accurate}, author = {Setiono, R.}, booktitle = {Neural Networks, IEEE Transactions on}, citeulike-article-id = {4774892}, citeulike-linkout-0 = {http://dx.doi.org/10.1109/72.839020}, citeulike-linkout-1 = {http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=839020}, doi = {10.1109/72.839020}, journal = {Neural Networks, IEEE Transactions on}, keywords = {neural-networks, phd, rule-extraction}, number = {2}, pages = {512--519}, posted-at = {2009-06-08 10:27:15}, priority = {2}, title = {Extracting M-of-N rules from trained neural networks}, url = {http://dx.doi.org/10.1109/72.839020}, volume = {11}, year = {2000} }

TY - JOUR ID - Setiono:2000 L3 - citeulike-article-id:4774892 N2 - An effective algorithm for extracting M-of-N rules from trained feedforward neural networks is proposed. First, we train a network where each input of the data can only have one of the two possible values, -1 or one. Next, we apply the hyperbolic tangent function to each connection from the input layer to the hidden layer of the network. By applying this squashing function, the activation values at the hidden units are effectively computed as the hyperbolic tangent (or the sigmoid) of the weighted inputs, where the weights have magnitudes that are equal one. By restricting the inputs and the weights to binary values either -1 or one, the extraction of M-of-N rules from the networks becomes trivial. We demonstrate the effectiveness of the proposed algorithm on several widely tested datasets. For datasets consisting of thousands of patterns with many attributes, the rules extracted by the algorithm are simple and accurate IS - 2 JF - Neural Networks, IEEE Transactions on EP - 519 TI - Extracting M-of-N rules from trained neural networks VL - 11 T2 - Neural Networks, IEEE Transactions on SP - 512 KW - neural-networks KW - phd KW - rule-extraction AU - Setiono, R PY - 2000/// UR - http://dx.doi.org/10.1109/72.839020 ER -