Predicting Actions from Induction on past Performance
Abstract
Human adversaries rely on knowledge about an opponent's previous performances in the adversarial domain. We simulate this aspect of human adversary performance by analyzing the records of prior encounters against specific adversaries. Our analysis acquires the perceptual chunks which form the foundations of an adversary's decision making ability. Induction is used to acquire high quality chunks which have been repeatedly used in prior adversarial encounters. Once the chunks for a specific adversary have been acquired, they are used to predict the future tactical and strategic decisions of that adversary. An implementation of the adversary modeling methodology is demonstrated for the domain of chess.
Cite
Text
Walczak. "Predicting Actions from Induction on past Performance." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50058-1Markdown
[Walczak. "Predicting Actions from Induction on past Performance." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/walczak1991icml-predicting/) doi:10.1016/B978-1-55860-200-7.50058-1BibTeX
@inproceedings{walczak1991icml-predicting,
title = {{Predicting Actions from Induction on past Performance}},
author = {Walczak, Steven},
booktitle = {International Conference on Machine Learning},
year = {1991},
pages = {275-279},
doi = {10.1016/B978-1-55860-200-7.50058-1},
url = {https://mlanthology.org/icml/1991/walczak1991icml-predicting/}
}