Playgol: Learning Programs Through Play

Authors: Andrew Cropper

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

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally test our claim on two domains: robot planning and real-world string transformations. Our experimental results suggest that playing can substantially improve learning performance.
Researcher Affiliation Academia Andrew Cropper University of Oxford andrew.cropper@cs.ox.ac.uk
Pseudocode Yes Algorithm 1 shows the Playgol algorithm
Open Source Code No The paper mentions that Playgol is implemented and uses Metagol (with a link to Metagol's GitHub), but it does not provide an explicit statement or link for the Playgol source code itself.
Open Datasets Yes Our dataset is based on the dataset from [Lin et al., 2014], which in turn is based on [Gulwani, 2011].
Dataset Splits No The paper specifies 5 training examples and 5 testing examples per task in the string transformation experiment, but does not mention a separate validation split or explicit percentages for the splits.
Hardware Specification No The paper mentions a timeout of 60 seconds per task but does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for the experiments.
Software Dependencies No The paper states 'Playgol uses Metagol [Cropper and Muggleton, 2016b]', but it does not specify a version number for Metagol or any other software dependencies.
Experiment Setup Yes For each p in {0, 200, 400, . . . , 2000}, we call playgol(Xn,BK,Tb,p,5) which returns a set of programs Pp. We enforce a timeout of 60 seconds per play and build task.