Scenarios to train and test results

[rpytel1]

Char based approach

• one-hot encoding config/char_based/one_hot.json
• 10 embedding layer dim config/char_based/10_dim.json
• 20 embedding layer dim config/char_based/20_dim.json

Try for both LSTM 1 layer and 2 layers (max 5/7 epochs)

Word-based approach

• 500 embedding layer dim config/word_based/500dim.json
• 100 embedding layer dim config/word_based/100dim.json
• 20 embedding layer dim 'config/word_based/20dim.json

Try for both LSTM 1 layer and 2 layers (max 5/7 epochs)

Code2vec training

• config/code_2_vec/code2vec.json (Still need some fixing with paths for dicts and data)

Code2vec pretrained output

• Linear (Casper)
• SVM (Jan)
• Random Forest (Jan)