Model saving and loading

[sethjuarez]

Add ability to save and load models

[civ101]

Any idea when the saving and loading a model will be complete? I have been able to save the model to file, but upon loading the file is when the model does not understand my programs classes. Otherwise this NuGet package has help save much time.

[sethjuarez]

I will work on it today and next week - sorry for the delay! I have been out of town a whole bunch. Would love some help!

[TonyHenrique]

Your Library is looking awesome, I hope you complete it and the examples and documentation too, in the same style of that Tennis example, for Supervised Learning and Unsupervised learning

[sethjuarez]

I added serialization to all of the supervised models (and tests for DT's and NaiveBayes). I also added helper methods on the base Model class to simplify adding serialization to models in the future.

[DashNY]

I was able to save a model into a file but I can't seem to figure out how to load it next time application starts.

SAVE:
learningModel = Learner.Learn(data, .80, 1000, generator);
learningModel.Model.Save(modelFileName);

LOAD: (not working, throwing NullRefException)
learningModel = new LearningModel();
learningModel.Model.Load(modelFileName);

[PippoAe]

I'm stuck on the same.
Saving works fine, but I don't know how to load it.

[rold2007]

@PippoAe I saw that recently @sethjuarez committed some code to better support serialization for NaiveBayes. Have you tried with the latest version ? Which generator did you use ? Have you seen this issue, it seems that after loading a model you still need to initialize the Descriptor.
If that doesn't solve your issue I would suggest you to open a new issue, including some code and steps to reproduce your issue.

[PippoAe]

Hi rold!
Thanks for the fast answer.
I now worked around the error.

I'm still using "model.Save(modelPath);" to save the TreeDecisionModel the learner has after learning.

To load I now use:
DecisionTreeModel model = (DecisionTreeModel)numl.Utils.Xml.Load(modelpath, typeof(DecisionTreeModel));

This returns me a working model again that I can use .Predict() on.

[sethjuarez]

Yay!!