basic model, at 15 epochs detects correct character with (conservatively) ~99.5% certainty run sample_synthesizer.py first to generate the samples you will be testing with