This documentation may require some updates as we have changed how model parameters are read
Should you wish to train model coefficients using Vowpal Wabbit please follow the instructions below.
You will need to have a Vowpal Wabbit source installation containing the script vw-varinfo. Set the environment as appropriate on your machine, e.g.,
verify the installation has the necessary script
run the training script
cd $ANTELOPE_TRAINING $ANTELOPE_DEMO/antelope/scripts/vw/train.sh
now inspect the model parameters
you should see something like the following (coefficients may evolve as we update the model by adding or changing parameters)
The values hardcoded in the class co.ifwe.antelope.bestbuy.exec.LearnedRankerScoring were fit using R. There is no need to change them now but you can replace them with the Vowpal Wabbit coefficients and expect to get very similar results.
Now head on back and follow the rest of the demo to test predictions made using the model.< Back to Documentation Index