Towards Human-centric AutoML via Logic and Argumentation

This repository contains detailed instructions on how to run the examples from the work Towards Human-centric AutoML via Logic and Argumentation.

Set up the environment

In this part we will prepare the environment required to run all the examples.

1. Install the Java runtime (version 11 or above). If Java is already present on your system, you can skip this step.
2. Download the stand-alone Arg2P Java IDE.
3. Double click on the file downloaded at step 2 (arg2p-ide-0.6.1-redist.jar). This should be the result. If nothing happens, check your Java installation by following the instructions at point 1.
4. You are now ready to run the example!

Example

1. Build the LogicalKB by typing the following theory.

t1 : [] => transformation(discretization).
t2 : [] => transformation(normalization).
a1 : [] => algorithm(dt).

c1 :=> mandatory_transformation_for_algorithm([discretization], dt).
c2 :=> invalid_transformation_set([normalization, discretization]).

% The following part is the engine that allows the argumation to be performed.
% It will be hidden, since will be part of the HAMLET framework.

g0 : algorithm(Z) => pipeline([], Z).
g1 : transformation(X), transformation(Y), algorithm(Z), prolog(X \== Y) =>
pipeline([X, Y], Z).
g2 : transformation(X), algorithm(Z) => pipeline([X], Z).

conflict([invalid_transformation_set(X)], [pipeline(Z, _)]) :-
\+ (member(Y, X), \+ member(Y, Z)).

conflict([mandatory_transformation_for_algorithm(X, Y)], [pipeline(Z, Y)]) :-
member(T, X),
\+ member(T, Z).

2. To require the evalution use the predicate buildLabelSets. Just put it in the query section (?-) and press the Solve button.