Abstract—A new modular recurrent neural network (MRNN)based speech-recognition method that can recognize the entire vocabulary of 1280 highly confusable Mandarin syllables is pro...
Learning from imbalanced data occurs frequently in many machine learning applications. One positive example to thousands of negative instances is common in scientific applications...
We demonstrate in this paper how certain forms of rule-based knowledge can be used to prestructure a neural network of normalized basis functions and give a probabilistic interpre...
Abstract. The EENCL algorithm [1] automatically designs neural network ensembles for classification, combining global evolution with local search based on gradient descent. Two mec...
Computer security, and intrusion detection in particular, has become increasingly important in today's business environment, to help ensure safe and trusted commerce between ...