save and load training result

May 8, 2009 at 7:53 AM

Is it possible to save and load the training result, so that it won't has be to train every time it determine a new entry?

Coordinator
May 12, 2009 at 3:43 PM

In the current version of the codebase, no, that code hasn't been written.  However, support for that is certainly planned.  Note how we're using a MemoryIndex ... one of the next things I work on is to provide a FileIndex, which would simply persist the contents of the memoryindex to disk every time an entry is added (along with support for loading from disk obviously).  Then, I also plan to have an additional assembly along with a sample schema implementation using a DatabaseIndex.  This implementation would obviously use the database to persist and retrieve stats.

So that's the plan, it should be relatively simple to do if you feel like providing your own, you just have to inherit from the Index class and make your own implementation if you can't wait for mine.  I'd love to get your thoughts on the design of the classes.

Thanks for checking out the library and thanks for the feedback :-)

Coordinator
Feb 19, 2010 at 6:17 PM

Just to follow up on this, the latest release has a FileIndex, which can be persisted and loaded from disk.