Learning to generate subgoals for action sequences

Summary form only given. None of the existing learning algorithms for neural networks in time-varying environments addresses the problems of learning to `divide and conquer´. Algorithms based on pure gradient descent or on adaptive critic methods are not suitable for dynamic control problems with long time lags between actions and consequences, and that there is a need for algorithms that perform `compositional learning´. The author discusses a system which solves at least one problem associated with compositional learning. The system learns to generate subgoals. This is done with the help of `time-bridging´ adaptive models that predict the effects of the system´s subprograms. An experiment on obstacle avoidance in a two-dimensional environment illustrates the approach