Minesweeper solver trained using machine learning

Requires numpy and sklearn to run

Pre-generated training data has been saved to disk via pickle, and can be found in the data/ folder.
If you wish to generate fresh training data:
    for the generation method of learning:
        run fivexfive_svm/fivexfive_block_generator.py
    for the gameplay method of learning:
        there are some lines in fivexfive_svm/fivexfive_svm_tester.py that need to be uncommented. These lines can be found by searching for "#Generate data as we go"
        then run the fivexfive_svm_tester.py until you are satisfied with the quantity of data generated.

If you're satisfied with the existing training data, and simply want to train the classifiers:
    fivexfive_svm/generation_svm_trainer.py will train the generation classifier
    fivexfive_svm/gameplay_svm_trainer.py will train the gameplay classifier.

The generation classifier is trained on randomly generated data.
The gameplay classifier is trained on data from real games the computer has already played. In theory, it should improve with each iteration of train -> generate new data -> train -> etc.

To see actual gameplay, use fivexfive_svm/fivexfive_svm_tester.py
    To change which classifier is tested, change the line commented "#choose your classifier here"

To view some statistics, run report_data/combined_validation.py

To see some baseline algorithms play, run the files in baseline_algorithms/

To play a game yourself, uncomment the last line in emulator/minesweeper_emulator.py and run it.