Types of Machine Learning Models: Explained with Examples

What Are the Different Types of Machine Learning Models?

Machine learning models can be categorized into three main types based on how they learn from data: supervised learning, unsupervised learning, and reinforcement learning. Each type is suited to specific tasks and has unique applications.

1. Supervised Learning

Definition: Supervised learning involves training a model on a labeled dataset, where each input has a corresponding output. The model learns to map inputs to outputs by minimizing the error between predictions and actual labels.

Common Use Cases:

• Classification: Predicting categories (e.g., spam detection in emails).
• Regression: Predicting continuous values (e.g., house price prediction).

Example in TensorFlow.js: Predicting house prices based on features like square footage and number of bedrooms.

import * as tf from '@tensorflow/tfjs';

// Dataset
const features = tf.tensor2d([[1000], [1500], [2000], [2500]]);
const labels = tf.tensor2d([[300000], [450000], [600000], [750000]]);

// Model
const model = tf.sequential();
model.add(tf.layers.dense({ inputShape: [1], units: 1 }));

model.compile({ optimizer: 'sgd', loss: 'meanSquaredError' });

// Train the model
model.fit(features, labels, { epochs: 50 }).then(() => {
  // Predict
  model.predict(tf.tensor2d([[1800]])).print();
});

2. Unsupervised Learning

Definition: Unsupervised learning deals with unlabeled data. The model tries to find patterns or structure in the data, such as grouping similar items or reducing dimensionality.

Common Use Cases:

• Clustering: Grouping similar data points (e.g., customer segmentation).
• Dimensionality Reduction: Reducing the number of features while retaining information (e.g., PCA for image compression).

Example in TensorFlow.js: Clustering customers based on purchasing behavior.

import * as tf from '@tensorflow/tfjs';

// Sample dataset
const data = tf.tensor2d([
  [1, 1],
  [2, 2],
  [3, 3],
  [10, 10],
  [11, 11],
  [12, 12],
]);

// Using k-means clustering with TensorFlow.js isn't built-in but can be implemented.
// Example: Centroid-based clustering
const centroids = tf.tensor2d([
  [2, 2],
  [11, 11],
]);

// Assign data points to clusters (concept demonstration)
data.sub(centroids.expandDims(0)).square().sum(-1).argMin(-1).print();

3. Reinforcement Learning

Definition: Reinforcement learning trains a model to make decisions by interacting with an environment. The model learns to maximize rewards by trying different actions and observing the outcomes.

Common Use Cases:

• Game Playing: Training an agent to play games like chess or Go.
• Robotics: Teaching robots to perform tasks in dynamic environments.

Example in TensorFlow.js: Training an agent to maximize score in a simple grid environment.

import * as tf from '@tensorflow/tfjs';

// Simplified reinforcement learning setup
// States: [0, 1, 2, 3] (positions on a grid)
// Actions: [left, right]
// Rewards: +1 for reaching state 3

const states = 4;
const actions = 2;
const qTable = tf.zeros([states, actions]);

// Simulate training: update Q-values
const learningRate = 0.1;
const discountFactor = 0.9;

const updateQValue = (state, action, reward, nextState) => {
  const currentQ = qTable.gather([state]).gather([action]);
  const maxNextQ = qTable.gather([nextState]).max();
  const newQ = currentQ.add(
    tf.scalar(learningRate).mul(
      tf.scalar(reward).add(tf.scalar(discountFactor).mul(maxNextQ)).sub(currentQ)
    )
  );
  qTable.scatterNDUpdate([[state, action]], [newQ]);
};