I would like to know what practices there are for using the vompal wabbit program (similar ones are possible, it doesn't matter here).
- Data is collected
- Trains model
- ?
- profit
It is not clear how to use the resulting model in a working environment.
After some searches, I found the following options:
- Running the vompal process interactively from a java application with the model received
- Running in daemon mode and communicating via sockets.
The first option is not sure that it is safe in terms of multithreading.
The second has a number of nuances. When sending data, you need some non-fixed time to wait (what?) For the results, after receiving the data it is not known whether it is worth waiting for a new piece of data or this is all. because data is a string and its format is impossible to understand where to finish reading.
Maybe there are some other solutions that are more stable and productive?
