This is a new framework to learn simples patterns from a few number of examples and use them to learn more complex ones. the learned models are perfectly interpretable and its depth depends on the question. What is meant by depth here is that whenever needed, the model learns to break down the problem into simpler subproblems and solves them using previously learned models, CDL is written in C++ and tested on 32 small datasets

The paper can be found at https://arxiv.org/abs/1605.00241

src/ - source files

dataset/ - three groups of small datasets (Group_A: 25 datasets, Group_B: 5 datasets,Group_C: 2 datasets). The first line in Group_A.dat contains the number of the small dataset in the file (must be 0). The next line contains the number of training sequences followed by the training seqences. The next line conatins number of testing sequences followed by the testing sequences.

experiments/

• 3 outputFiles: contain MGICDs learned in each group of datasets and some details about training
• 32 animation files that explain how MGICD solve the test questions, they can be shown by running animate.py but you need first to change the directory of the animation file in the first line in animate.py. You can also change the animation speed by changing the value of nodeTime (Tested on Python 3).
• Most of PCDs in the outputFiles can be displayed by visualizeCD.py, but it need first to be copied in CD.dat

$ make
$ ./runner

You can edit Driver.cpp to learn datasets in Group_A, Group_B or Group_C. For example, to learn the first dataset in Group_A write this in main() function:

    IOH.loadDatasetsAndQuestionsFromFile("dataset/Group_A.dat");
    Constants::redirectStdoutToOutputFile = true;  //redirect stdout to OutputFile.txt
    setHyperParameters(1,2,1,2,7,0,0,1);           //set hyperparamters for the 25 dataset in file Group_A.dat
    TE.learnAndTest(0);                            //learn & test the first Dataset in file Group_A.dat

The results and learned models will be saved in OutputFile.txt. To learn all the datasets in Group_A:

    IOH.loadDatasetsAndQuestionsFromFile("dataset/Group_A.dat");
    Constants::redirectStdoutToOutputFile = true;  //redirect stdout to OutputFile.txt
    setHyperParameters(1,2,1,2,7,0,0,1);           //set hyperparamters for the 25 dataset in file Group_A.dat
    TE.learnAndTest();                             //learn & test all Datasets in file Group_A.dat

You can also write a new dataset and test it but it must be formatted like other datasets

This project is licensed under the terms of the MIT license