Members of AIC succeeded recently in highly selective paper review process of the top Artificial Intelligence conference AAAI-19. Although this year rejection rate was roughly 16%, AIC got accepted 4 papers in various topics of game theory, automated planning and security.
Congratulation to Daniel Fišer, Jaromír Janisch, Karel Horák and Lukáš Chrpa
1, Jaromír Janisch, Tomáš Pevný, Viliam Lisý
Název: Classification with Costly Features using Deep Reinforcement Learning
Anotace: Cost is an integral part of many real-world problems, let it be medicine, insurance or computer security. In each of these domains, any piece of information about a sample at hand can be retrieved only with some effort and after spending valuable resources - time, money, computational power, etc. Our work focuses on a classification problem under these constraints and tries to find an accuracy-cost balance. We employ recent insights from Deep Reinforcement Learning and present a robust, flexible and accurate technique to deal with this problem.
Odkaz na článek: https://arxiv.org/abs/1711.07364
Odkaz na zdrojové kódy: https://github.com/jaromiru/cwcf
2,Lukas Chrpa, Mauro Vallati (Univ. of Huddersfield):
Title: Improving Domain-independent Planning via Critical Section
Abstract: Macro-operators, macros for short, are a well-known technique for enhancing performance of planning engines by providing ``short-cuts'' in the state space. Existing macro learning systems usually generate macros from most frequent sequences of actions in training plans. Such approach priorities frequently used sequences of actions over meaningful activities to be performed for solving planning tasks.
This paper presents a technique that, inspired by resource locking in critical sections in parallel computing, learns macros capturing activities in which a limited resource (e.g., a robotic hand) is used. In particular, such macros capture the whole activity in which the resource is ``locked'' (e.g., the robotic hand is holding an object) and thus ``bridge'' states in which the resource is locked and cannot be used. We also introduce an ``aggressive'' variant of our technique that removes original operators superseded by macros from the domain model. Usefulness of macros is evaluated on several state-of-the-art
planners, and a wide range of benchmarks from the learning tracks of the 2008 and 2011 editions of the International Planning Competition.
3,Daniel Fišer, Álvaro Torralba, Alexander Shleyfman
Title:Operator Mutexes and Symmetries for Simplifying Planning Tasks
4, Karel Horák, Branislav Bošanský
Název: Solving Partially Observable Stochastic Games with Public Observations
Anotace: Dynamic interaction and imperfect perception of situations are fundamental components of many real-world security problems. In order to design strong defensive strategies, computing near-optimal strategies in infinite scenarios with imperfect information is critical. Formally, such scenarios can be described using partially observable stochastic games (POSGs). We focus on a subclass of such models, POSGs with public observations, and provide a practical algorithm for solving such games.