Learning Symbolic Timed Models from Concrete Timed Data

Dierl, Simon; Howar, Falk; Kauffman, Sean; Kristjansen, Martin; Larsen, Kim Guldstrand; Lorber, Florian; Mauritz, Malte

doi:10.1007/978-3-031-33170-1_7

Cited by 2 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several works exist to make the decision-making of black-box systems used for control explainable. Automata learning refers to techniques that infer a surrogate model (e.g., in the form of an input-output automaton [41], a timed automaton [13] or an MDP [38]) from a given black-box system by observing its behavior. The tool dtControl [5] learns decision trees for hybrid and probabilistic control systems, and has been recently extended to support richer algebraic predicates as splitting rules with the use of support vector machines [22].…”

Section: Related Workmentioning

confidence: 99%

Formal XAI via Syntax-Guided Synthesis

Bjørner,

Judson,

Cano

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

In this paper, we propose a novel application of syntaxguided synthesis to find symbolic representations of a model's decisionmaking process, designed for easy comprehension and validation by humans. Our approach takes input-output samples from complex machine learning models, such as deep neural networks, and automatically derives interpretable mimic programs. A mimic program precisely imitates the behavior of an opaque model over the provided data. We discuss various types of grammars that are well-suited for computing mimic programs for tabular and image input data. Our experiments demonstrate the potential of the proposed method: we successfully synthesized mimic programs for neural networks trained on the MNIST and the Pima Indians diabetes data sets. All experiments were performed using the SMT-based cvc5 synthesis tool.

show abstract