next up previous
Next: Bibliography Up: PLANNING SIMPLE TRAJECTORIES USING Previous: LIMITATIONS

ACKNOWLEDGEMENTS

Thanks to Mike Mozer for helpful comments on a draft of this paper. This research was supported in part by a DFG fellowship to J. Schmidhuber, as well as by NSF PYI award IRI-9058450, grant 90-21 from the James S. McDonnell Foundation, and DEC external research grant 1250 to Michael C. Mozer.

Figure 4: In this example, the task is to find a trajectory (composed of three `sub-trajectories') leading from START to GOAL. The big circle represents a single swamp. The evolution of two subgoals emitted by an adaptive recurrent subgoal generator is shown. Check out Schmidhuber's Habilitation thesis for pictures.

Figure 5: Here many swamps separate the START location from the GOAL location. The evolution of five subgoals (represented by little black dots) emitted by a non-recurrent subgoal generator is shown. Check out Schmidhuber's Habilitation thesis for pictures.



Juergen Schmidhuber 2003-03-14

Back to Subgoal learning - Hierarchical Learning
Pages with Subgoal learning pictures