Neural-Symbolic Learner for Michalski's train problem

• open and run Neural_Symbolic_Learner.ipynb on google colab or localy with jupyter notebook
  • the main dependencies are shown on the first cell to be run. It installs tensorflow, keras and other modules that are used in the notebook.
  • run all cells to get the result. It may take a while (estimated time: 2 hours).

Michalski’s train details

• The problem:

  • there are ten railway trains;
  • five are travelling east;
  • five are travelling west;
  • each train comprises a locomotive pulling wagons;

• Whether a particular train is travelling towards the east or towards the west is determined by some properties of that train:

  • appended wagons;
  • short or long wagons;
  • closed or open wagons;
  • jagged or not;
  • shaped the wagons contain;
  • how many shapes;
  • and more.

• The data describes different features of trains. The positive examples are the trains on the left in the figure below and the negative examples are the trains on the right:

[Fig. 1]

The task of the learner is to find characteristics that determine the direction the train take.