LBAQ: a pattern recognition neural network that learns by asking questions

LBAQ (learning by asking questions) is a fast learning neural architecture that forms internal representations of the given examples and asks the teacher about labels for each of them. A considerable reduction in the length of the training session is achieved with LBAQ on the order of 1:500 over the time needed by a backpropagation network. The simulation performance results of this network on standard problems are superior to those for backpropagation networks in terms of ease, speed of training, and the ability to incrementally train the network on subjects of the training set at different times as opposed to the lengthy one-shot training session in the backpropagation case.<
>