Active Learning Algorithms for Neural Networks

......_.ad- Five neural algorithms are described that have been derived from an incremental Je..iDg framew.prk, called GENIAL. The GENIAL learning employs four learning elements Ia adapt the network structure and weights while exploring its environment to acquire...el. informatiotl. The common feature of all the algorithms is the active use of training.... _ This includes scheduling the presentation order of given acamples, selecting a critical s-llset of a large data set, and generating new examples. The algorithms are described in die..der of increasing activeness or autonomy and compared against four different tasks. 'l.'lle results show improved generalization performance as the autonomy of the algorithm