mlt-node

Node port of the BYU CS 478 machine learning toolkit written in typescript

Getting Started

1. Install mlt-node

``bash
npm install mlt-node
`
2. Download some datasets
`bash
mkdir datasets
wget http://axon.cs.byu.edu/~martinez/classes/478/stuff/iris.arff -P datasets/
`
3. Write a program to take in parameters and call the toolkit. This can be as simple as:
`typescript
import { SupervisedLearner, BaselineLearner, run } from 'mlt-node';

function getLearner(model: string): SupervisedLearner {
switch (model) {
case 'baseline':
return new BaselineLearner();
case 'perceptron':
// return new Perceptron();
case 'neuralnet':
// return new NeuralNet();
case 'decisiontree':
// return new DecisionTree();
case 'knn':
// return new InstanceBasedLearner();
default:
throw new Error('Unrecognized model: ' + model);
}
}

//Parse the command line arguments
const learnerName = process.argv[3];

run(getLearner(learnerName));
`

4. Compile your typscript program and run
`bash
node compiledProgram.js -L baseline -A datasets/iris.arff -E training
`

Creating Learners

Creating new learners is as simple as extending the SupervisedLearner class provided by the toolkit. Just make sure to override the train() and predict() functions of the SupervisedLearner base class.

`typescript
import { Matrix, SupervisedLearner } from 'mlt-node';

class MyNewLearner extends SupervisedLearner {

train(features: Matrix, labels: Matrix) {
// Your training algorithm here.
}

predict(features: number[], labels: number[]) {
// Your prediction algorithm here.
}
}
``

Contributing

There are bound to be bugs in this project. Please help fix them by creating PRs.