Partitioned Least Squares

This repository contains the code to replicate experiments on the PartitioningLS model as reported in the article published on TODO.

Setup you environment

Clone the repository and move into the project directory.

From inside the project directory update the Julia packages by running:

julia --project=. --color=yes -e 'using Pkg; Pkg.update()'

Download the datasets

The datasets directory contains one subfolder for each available dataset.

In each folder you can find either:

• the actual data, in this case you will see files data.csv and blocks.csv already present in the directory
• or a script to generate the data. In this case you will find a CreateData.jl file, that you can launch as it follows:

    julia --project=../.. CreateData.jl

• or scripts two download and preprocess the data. In this case you can do these operations as it follows:

    bash download.sh
    julia --project=../.. convert.jl 

All five datasets need to be downloaded before the experiments can be run. If you fail to download them, or you don't want to use some of them, you can modify the run_* scripts to avoid experimenting with them, but note that the scripts that will allow you to format the results require all results to be present. Updating them is feasible, but not as easy as changing the bash scripts.

Run the experiments

bash run_model_quality_experiments.sh
bash run_time_vs_obj_experiments.sh

The analysis of the Housing dataset is included in a julia notebook that can be found in the housing directory.

Accessing results

After running the experiments, the results will be stored in the experiments directory.

You can get a summary table of the results about model-quality by running:

julia --project=. src/CompareResults.jl

You can create a pdf file with a plot of the time-vs-obj experiments by running:

pdflatex src/time-vs-obj.tex

and access the results by opening the time-vs-obj.pdf file.

Citing

If you use this code in your research, please cite the following article:

TODO