Density-Adaptive Learning and Forgetting
Abstract
We describe a density-adaptive reinforcement learning and a density-adaptive forgetting algorithm. This learning algorithm uses hybrid Dκ-D/2κ -trees to allow for a variable resolution partitioning and labelling of the input space. The density adaptive forgetting algorithm deletes observations from the learning set depending on whether subsequent evidence is available in a local region of the parameter space. The algorithms are demonstrated in a simulation for learning feasible robotic grasp approach directions and orientations and then adapting to subsequent mechanical failures in the gripper.
Cite
Text
Salganicoff. "Density-Adaptive Learning and Forgetting." International Conference on Machine Learning, 1993. doi:10.1016/B978-1-55860-307-3.50042-3Markdown
[Salganicoff. "Density-Adaptive Learning and Forgetting." International Conference on Machine Learning, 1993.](https://mlanthology.org/icml/1993/salganicoff1993icml-density/) doi:10.1016/B978-1-55860-307-3.50042-3BibTeX
@inproceedings{salganicoff1993icml-density,
title = {{Density-Adaptive Learning and Forgetting}},
author = {Salganicoff, Marcos},
booktitle = {International Conference on Machine Learning},
year = {1993},
pages = {276-283},
doi = {10.1016/B978-1-55860-307-3.50042-3},
url = {https://mlanthology.org/icml/1993/salganicoff1993icml-density/}
}