Learning by Analyzing Fortuitous Occurrences
Abstract
In complex, real-world domains, complete modelling of the world is a computationally intractable task. One difficulty with planning in such situations is that inefficient, suboptimal plans are generated. This is because determining the best way to achieve a goal may depend on many factors. This paper presents methods for detecting and improving plans which achieve the goal but in a suboptimal manner. By noticing and analyzing fortuitous occurrences, the system can improve its plans and hence its performance. These methods cover both learning from observation and from the system's own problem-solving and represent one portion of a general framework of refinement for inference-limited systems.
Cite
Text
Chien. "Learning by Analyzing Fortuitous Occurrences." International Conference on Machine Learning, 1989. doi:10.1016/B978-1-55860-036-2.50071-0Markdown
[Chien. "Learning by Analyzing Fortuitous Occurrences." International Conference on Machine Learning, 1989.](https://mlanthology.org/icml/1989/chien1989icml-learning/) doi:10.1016/B978-1-55860-036-2.50071-0BibTeX
@inproceedings{chien1989icml-learning,
title = {{Learning by Analyzing Fortuitous Occurrences}},
author = {Chien, Steve A.},
booktitle = {International Conference on Machine Learning},
year = {1989},
pages = {249-251},
doi = {10.1016/B978-1-55860-036-2.50071-0},
url = {https://mlanthology.org/icml/1989/chien1989icml-learning/}
}