search

mdabros / SharpLearning

Machine learning for C# .Net
MIT License
384 stars 85 forks source link

retrain a Model #122

Closed kiranmaya closed 5 years ago

kiranmaya commented 5 years ago

How to Retain a model with new data ?

mdabros commented 5 years ago

@kiranmaya If you simply want to retrain the model from scratch on new data, you can simply re-run you training code again, pointing to the new data. If by retrain, you mean continue training on an existing model, this is currently not possible.

kiranmaya commented 5 years ago

@mdabros ya read about it ,some algos needs all the data at once ,like xgboost regression .

mdabros commented 5 years ago

@kiranmaya Yes, exactly. Models based on iterative optimization, like stochastic gradient descent for neural networks, can be used for continuous learning, and here it is possible to continue training on an existing model. The neural network implementation in SharpLearning sadly lacks this feature currently though.

mdabros commented 5 years ago

@kiranmaya I am closing this issue since an answer has been provided.