Artificial Neural Network Classification

Using a labelled training set, an ANN classifier can be trained using supervised learning algorithms to discriminate between normal and abnormal regions. Features from representative training samples are provided during supervised learning and the weights of the ANN are updated until the generalization ability of the classifier starts to decrease measured on a separate validation set. Implementation in the adaptive knowledge-based model results in the construction of a separate ANN classifier for each breast type grouping. Only regions from mammograms of the same mammogram type will be considered for each ANN. Each ANN is a three-layer feed-forward network comprised of a different number of hidden nodes and two output nodes (normal, abnormal). The optimal number of hidden nodes is determined for each ANN individually. To ensure an unbiased result and that every sample is used at least once in training and testing, a 10-fold cross validation strategy [32] is employed. No sample appears simultaneously in training and test. Additionally a validation set is used (comprising 10% of the training samples) to prevent over-fitting of the ANN to the training set. The feed-forward ANN is trained using a back-propagation with momentum learning function (learning rate n = 0.01, momentum ¡x = 0.5) together with a softmax activation function and used on test to give an estimate of the a posteriori probability of each pattern for each class.

0 0

Post a comment