TensorForge Guide

TensorForge is a high-performance TypeScript machine learning framework featuring raw Float64 typed structures, direct analytical utility sets, and end-to-end model tooling.

Installation

git clone https://github.com/philipszdavido/TensorForgejs.git
cd TensorForgejs
npm install
npm run build

Recommended Architecture Workflow

1. Format raw mathematical primitives via dedicated structures.
2. Instantiate target model classes with structural initializers.
3. Execute a matrix/vector forward pass to build inferences.
4. Pass structural evaluations into target loss classes.
5. Propagate backward gradients straight through graph links.
6. Update parameters via local hooks or custom optimizers.

Import Specifications

import TensorForge from 'tensorforgejs';
const { Core, Models, ModelInferenceEngine } = TensorForge;

// Or resolve individual explicit symbols
import { Matrix, Vector, Tensor } from 'tensorforgejs';
import { LinearRegression, KNN, NeuralNetwork } from 'tensorforgejs';

Production Pipeline Loop

const model = new Models.LinearRegression(inputSize);
const learningRate = 0.01;

for (let epoch = 0; epoch < epochs; epoch++) {
  for (const sample of trainingData) {
    const yHat = model.predict(sample.x);
    model.backward(sample.x, sample.y, yHat);
    for (let i = 0; i < model.weights.length; i++) {
      model.weights[i] -= learningRate * model.gradWeights[i];
    }
    model.bias -= learningRate * model.gradBias;
  }
}