| forum.alglib.net http://forum.alglib.net/ |
|
| Saving trained classifier http://forum.alglib.net/viewtopic.php?f=2&t=297 |
Page 1 of 1 |
| Author: | scorelord [ Fri Feb 11, 2011 12:36 am ] |
| Post subject: | Saving trained classifier |
Hi, I really like Alglib and I enjoy using its many functions. One of the things I'm looking into is saving a trained classifier. I use the random forest classifier quite extensively. However, the one thing I do not like about it is the required training time which is much longer than the classification time in my application. I was wondering if it is possible to somehow save the trained classifier to disk so I could just load it straight into memory and start classifying. I'd appreciate any tips you could give me. Thanks Mr. Bochkanov for your wonderful library.... |
|
| Author: | Sergey.Bochkanov [ Mon Feb 14, 2011 1:38 pm ] |
| Post subject: | Re: Saving trained classifier |
It can be problematic - as of 3.2, ALGLIB doesn't support serialization of complex data structures. Serialization was planned for a 3.3, but I suspect that it will wait for 3.4 at best. |
|
| Author: | efimovandr [ Sun Feb 20, 2011 8:52 pm ] |
| Post subject: | Re: Saving trained classifier |
Hi. In my application I use the neural network from C# and I have to store a trained networks. I did a copy of the alglib mlpbase.multilayerperceptron (with litle changes). Then I marked all fields with xml serialization attributes. And now I can serialize my network and store it in the any type of storage. Before using this lib you should create a good adapter I think. |
|
| Page 1 of 1 | All times are UTC |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|