Compact Representation of Value Function in Partially Observable Stochastic Games

Authors: Karel Horák, Branislav Bošanský, Christopher Kiekintveld, Charles Kamhoua

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluation confirms that the new algorithm using the compact representation dramatically increases scalability compared to the state of the art. [...] 5 Experimental Results In this section we experimentally evaluate the scalability and properties of our proposed abstraction technique based on the model from Section 4. [...] Figure 2 shows the runtimes of the original algorithm applied to the unabstracted game and our proposed approach. [...] In Table 1, we present the empirical bound on the relative distance from the equilibrium of the unabstracted bound.
Researcher Affiliation Collaboration Karel Hor ak1 , Branislav Boˇsansk y1 , Christopher Kiekintveld2 and Charles Kamhoua3 1Czech Technical University in Prague, FEE, Department of Computer Science 2The University of Texas at El Paso, Computer Science Department 3Army Research Laboratory, Network Security Branch
Pseudocode Yes Algorithm 1: HSVI algorithm for OS-POSGs when summarized abstraction is used.
Open Source Code No The paper does not provide a link or explicit statement about the availability of open-source code for the described methodology.
Open Datasets No We used a set of randomly generated graphs (varying the number of vertices) and we attempted to solve these instances using both the original (unabstracted) approach and the algorithm we present in this paper.
Dataset Splits No The paper does not specify exact percentages or sample counts for training, validation, or test dataset splits.
Hardware Specification Yes All of the experiments were run on a machine with an Intel i7-8700K and 32GB of RAM.
Software Dependencies Yes We used CPLEX 12.7.1 to solve the linear programs used in the algorithms.
Experiment Setup Yes The parameters used for the original algorithm follow the parameters proposed in [Hor ak et al., 2017]. We modified the initialization of the bounds to make it valid for the undiscounted problem in question. The target precision ϵ = 0.1 was used for both algorithms.