Quickstart Go

Go Quickstart

Lemkit for Go currently implements only reading previously saved linear classification models (the equivalent of Scala lemkit-model); see Quickstart Scala. There are command-line applications to do prediction, and this can also be done using library functions.

Library functions

This library assumes you have already trained a model using Scala or Python. For the following, we assume you have a binary model saved in model.binary, but the code is very similar for JSON models. We also assume you have a file data.predict containing data instances to predict, using the format described in Input Format.

Example code is as follows. See also github.com/peoplepattern/lemkit/go/lkpredict/PredictApp, which implements lkpredict.

package main

import (
    "fmt"
    "github.com/peoplepattern/lemkit/go/classify"
)

func main() {
    classifier, err := classify.ReadBinaryModel("model.binary")
    if err != nil {
        fmt.Println(err)
        os.Exit(1)
    }

    predictData, err := classify.ReadDataFile("data.predict")
    if err != nil {
        fmt.Println(err)
        os.Exit(1)
    }

    for i := 0; i < len(predictData); i++ {
        prediction, _ := classifier.Predict(predictData[i].Features)
        correct := predictData[i].Label
        fmt.Printf("%d %s %s\n", i+1, correct, prediction)
    }
}

Note that the second return value to classifier.Predict is a probability. If you want the full set of scores for each possible label, use classifier.Scores or classifier.ScoresMap. (These are methods of the LinearClassifier struct.)

If you want to read a JSON model instead of a binary model, use classify.ReadJSONModel.

Command-line applications

lkpredict

lkpredict is used to do prediction on linear classifier models trained using Scala or Python. The following command does the equivalent of the library code above:

  lkpredict -m vowpal -f binary -p data.predict

Full documentation is available in the User Guide.