juditacs / semeval

MathLing Budapest Team's repo
MIT License
10 stars 9 forks source link

Dev2 #54

Closed pajkossy closed 9 years ago

pajkossy commented 9 years ago

@recski featurizing and training can be done separately, the table containing the featurized training and test data are stored in RegressionModel, which can be dumped and later reloaded. There are also some new options for choosing the learning algorithm, these are imported from sklearn. There is a new section ('ml') in the config files (see regression.cfg) for specifying if the featurized training data is to be dumped (and where), to load(from where), and which training algorithm to choose.