optparse

[zseder]

it would be easier to run experiments without always changing parameters (and more git friendly)

[zseder]

too much option, probably a config file would be better (yaml? python cfg?)

• training_file, mandatory
• test_file, optional
• valid_file, optional
• train_split, optional, only used, when test and valid files are both absent
• window_size, optional, with default value
• embedding size, optional, with default value
• hidden layer size, optional, with default value (there will be more layers later, so this is a temporary solution)
• training_epochs, optional, with default value
• regularization, optional, with default value (0 or 1e-5)
• momentum, optional, False by default
• momentum_saturation, optional
• adjust_learning_rate, optional, False by default
• start_learning_rate, optional
• learning_rate_decay, optional
• learning_rate_saturation, optional

[zseder]

solved in #26