Forward Modeling for Partial Observation Strategy Games - A StarCraft Defogger

Authors: Gabriel Synnaeve, Zeming Lin, Jonas Gehring, Dan Gant, Vegard Mella, Vasil Khalidov, Nicolas Carion, Nicolas Usunier

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments We hypothesize that our models will be able to make predictions of the global build tree and local unit numbers better than strong existing baselines. We evaluate our models on a human games dataset, and compare them to heuristic baselines that are currently employed by competitive rule-based bots. We also test whether we are able to use the defogger directly in a state-of-the-art rule-based Star Craft bot, and we evaluate the impact of the best models within full games.
Researcher Affiliation Industry Gabriel Synnaeve Facebook, NYC gab@fb.com Zeming Lin Facebook, NYC zlin@fb.com Jonas Gehring Facebook, Paris jgehring@fb.com Dan Gant Facebook, NYC danielgant@fb.com Vegard Mella Facebook, Paris vegardmella@fb.com Vasil Khalidov Facebook, Paris vkhalidov@fb.com Nicolas Carion Facebook, Paris alcinos@fb.com Nicolas Usunier Facebook, Paris usunier@fb.com
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes We release the code necessary for the reproduction of the project at https://github.com/facebookresearch/starcraft_defogger.
Open Datasets Yes Our models are trained and evaluated on the STARDATA corpus, which consists of 65,000 high quality human games of Star Craft: Brood War [19].
Dataset Splits Yes We use the train, valid, and test splits given by the authors, which comprise 59060, 3289 and 3297 games, respectively.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No Our models are implemented in Py Torch [21], and data preprocessing is done with Torch Craft [22]. (While tools are mentioned, specific version numbers for these software dependencies are not provided.)
Experiment Setup No We explored the following properties: kernel width of convolutions and striding (3,5); model depth; non-linearities (ReLU, GLU); residual connections; skip connections in the encoder LSTM; optimizers (Adam, SGD); learning rates. Please check the appendix for a more detailed description of hyperparameters searched over.