TY - GEN
T1 - SOS
T2 - 16th International Conference on Quantitative Evaluation of Systems, QEST 2019
AU - Ashok, Pranav
AU - Křetínský, Jan
AU - Larsen, Kim Guldstrand
AU - Le Coënt, Adrien
AU - Taankvist, Jakob Haahr
AU - Weininger, Maximilian
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - For hybrid Markov decision processes, Stratego can compute strategies that are safe for a given safety property and (in the limit) optimal for a given cost function. Unfortunately, these strategies cannot be exported easily since they are computed as a very long list. In this paper, we demonstrate methods to learn compact representations of the strategies in the form of decision trees. These decision trees are much smaller, more understandable, and can easily be exported as code that can be loaded into embedded systems. Despite the size compression and actual differences to the original strategy, we provide guarantees on both safety and optimality of the decision-tree strategy. On the top, we show how to obtain yet smaller representations, which are still guaranteed safe, but achieve a desired trade-off between size and optimality.
AB - For hybrid Markov decision processes, Stratego can compute strategies that are safe for a given safety property and (in the limit) optimal for a given cost function. Unfortunately, these strategies cannot be exported easily since they are computed as a very long list. In this paper, we demonstrate methods to learn compact representations of the strategies in the form of decision trees. These decision trees are much smaller, more understandable, and can easily be exported as code that can be loaded into embedded systems. Despite the size compression and actual differences to the original strategy, we provide guarantees on both safety and optimality of the decision-tree strategy. On the top, we show how to obtain yet smaller representations, which are still guaranteed safe, but achieve a desired trade-off between size and optimality.
UR - http://www.scopus.com/inward/record.url?scp=85072854193&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-30281-8_9
DO - 10.1007/978-3-030-30281-8_9
M3 - Conference contribution
AN - SCOPUS:85072854193
SN - 9783030302801
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 147
EP - 164
BT - Quantitative Evaluation of Systems - 16th International Conference, QEST 2019, Proceedings
A2 - Parker, David
A2 - Wolf, Verena
PB - Springer Verlag
Y2 - 10 September 2019 through 12 September 2019
ER -