Abstraction, reformulation and approximation

Más información sobre el libro

This collection features a range of papers and research summaries on various aspects of program verification, abstraction methods, and heuristic search in stochastic environments. Topics include generating admissible heuristics, synthesizing plans across multiple domains, and implementing frameworks for soft constraints. The work explores learning regular expressions from noisy sequences and transitioning from hierarchical hidden Markov models to Bayesian networks. It revisits hierarchical heuristic search and presents models for sequence and text classification using multinomial event abstractions. The collection also delves into polynomial reachability checking for Petri nets and symmetry detection in problem specifications. Additional studies focus on approximate model-based diagnosis, feature selection, and construction to enhance concept learning in high-dimensional data. The papers address qualitative spatio-temporal abstractions in disaster scenarios and automatic concept formation through minimum description length. The research includes experiments with multiple abstraction heuristics in symbolic verification and probabilistic abstraction for uncertain temporal data. It also covers knowledge acquisition in flow and water quality models, state abstraction for real-time video games, and learning skills in reinforcement learning through relative novelty. Invited talks discuss efficient SQL query responses and abstract r