Companion repository for the 2020 article "The Predictive Power of Spatial Relational Reasoning Models: A New Evaluation Approach" by Marco Ragni, Daniel Brand, and Nicolas Riesterer.
analysis/
: Contains the analysis scripts and corresponding datasetsbenchmarks/
: Contains the CCOBRA benchmark configuration files.data/
: Contains the datasets.models/
: Contains the CCOBRA model scripts.*.json
: CCOBRA benchmark configuration files.*.csv
: CCOBRA benchmark result files.
visualization/
: Contains the visualization scripts.results/
: Contains the resulting image files.heatmap.py
: Plots the optimality comparison heatmap (Figure 6).performance.py
: Plots the performance boxplots (Figures 2-5).
- Python 3
Navigate into the analysis/benchmarks/
directory and run the respective benchmark .json
file via CCOBRA:
$> cd /path/to/repository/analysis/benchmarks
$> ccobra 3ps.json -s 3ps.csv
The command line option -s <filename>
stores the results directly in a CSV file.
Navigate into the analysis/visualization/
directory and run the performance.py
script:
$> cd /path/to/repository/analysis/visualization
$> python performance.py ../benchmarks/3ps.csv
Navigate into the analysis/visualization/
directory and run the heatmap.py
script:
$> cd /path/to/repository/analysis/visualization
$> python heatmap.py