aluriak / concept-generation

Comparison and testing of multiple concept generation method using ASP

Concept generation comparison

Multiple concept generation methods are implemented in ASP, and compared in runtime for a randomly generated set of contexts of various size and density, and also proved to work by testing against controlled and verified sets of contexts.

Results

Here is the main result of this repository: showing that all methods are roughly equivalent.

The false method is a reference: a method that do nothing, and thus expected to fail (this is verified by unit tests).

Features

Running the tests

Verify that all methods find correctly the expected solutions/concepts in hand-crafted contexts.

make tests

Should be all green.

Running the benchmark

Extensive computation. Will melt your CPU and generate the output.csv file.

make benchmark

Should be all white.

Plotting the results

Using the results of the benchmark, allow one to visualize the results. You should hack around the drawing module to get exactly what you want (by default it just look for measures about contexts with a density of 0.4).

make draw

Should be all fighting for the first place.

Requirements

See requirements.

Method detection

See methods module to see the methods automatic and static generation.