Introduction to Machine Learning

Machine Learning (ML) is one of the most important fields in modern computer science and artificial intelligence. It enables systems to learn from data instead of being explicitly programmed with rules. This first lecture in the Core Machine Learning series sets the stage for deeper exploration of algorithms, math foundations, and applications.

---

1. What is Machine Learning?

Arthur Samuel defined ML in 1959 as:

“The field of study that gives computers the ability to learn without being explicitly programmed.”

Tom Mitchell gave a more formal definition in 1997:

“A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.”

Key components of ML definition:

• Task (T): What problem is the algorithm solving? (e.g., classify emails as spam or not).
• Experience (E): The data it learns from (e.g., labeled emails).
• Performance Measure (P): How success is measured (e.g., classification accuracy).

What ML is NOT

• It is not magic – it requires quality data and good features.
• It is not rule-based programming – no fixed if-else rules.
• It is not human-level intelligence – though it powers parts of AI.

---

2. Why is Machine Learning Important?

ML underpins nearly every modern AI system:

• Recommendation systems (YouTube, Netflix, Amazon).
• Search engines (Google uses ML ranking).
• Speech recognition & NLP (Siri, Alexa, GPT).
• Computer vision (self-driving cars, medical imaging).
• Fraud detection (banking, cybersecurity).

The Explosion of ML

Three factors caused ML’s rapid growth:

1. Data availability (Big Data, IoT).
2. Computational power (GPUs, TPUs, cloud).
3. Algorithmic advances (deep learning, transformers).

---

3. Types of Machine Learning

3.1 Supervised Learning

• Learns from labeled data (inputs + correct outputs).
• Goal: Predict output for new unseen inputs.
• Examples: Linear regression, logistic regression, decision trees, SVMs, neural networks.

Formula (Regression):

$$y=f(x)+ϵ$$

where $f(x)$ is the learned function, $y$ is the output, and $ϵ$ is noise.

Explanation

Supervised learning assumes a mapping from features $x$ to labels $y$. The model generalizes this mapping from known examples.

3.2 Unsupervised Learning

• Learns from unlabeled data (only inputs, no outputs).
• Goal: Find hidden structure in data.
• Examples: Clustering (k-means), dimensionality reduction (PCA).

3.3 Reinforcement Learning (RL)

• Learns by interacting with an environment.
• Receives rewards or penalties as feedback.
• Examples: AlphaGo, robotics, recommendation personalization.

3.4 Semi-Supervised & Self-Supervised Learning

• Semi-supervised: Mix of labeled and unlabeled data.
• Self-supervised: Learns signals from data itself (e.g., predicting missing words → GPT training).

---

4. Typical ML Workflow

1. Problem Definition – What is being predicted?
2. Data Collection – Gather data (structured/unstructured).
3. Data Preprocessing – Cleaning, handling missing values, normalization.
4. Feature Engineering – Selecting or creating informative variables.
5. Model Selection – Choosing appropriate algorithms.
6. Training – Optimizing model parameters on training data.
7. Evaluation – Testing on unseen data, using metrics.
8. Deployment – Serving models in production.
9. Monitoring & Maintenance – Detecting drift, updating models.

---

5. The Role of Mathematics in ML

ML is powered by math, especially:

• Linear Algebra → Representing data (vectors, matrices).
• Calculus → Optimization, gradients, backpropagation.
• Probability & Statistics → Modeling uncertainty.
• Information Theory → Loss functions, entropy, KL divergence.

Why math matters

Without math, ML becomes a black box. Understanding math ensures interpretability, debugging ability, and better research contributions.

---

6. Minimal Code Example (Python & Rust)

Python

from sklearn.linear_model import LinearRegression
import numpy as np

# dataset
X = np.array([[1], [2], [3], [4], [5]])
y = np.array([2.2, 4.1, 6.0, 8.1, 9.9])

# train model
model = LinearRegression().fit(X, y)

# prediction
print("Prediction for 6:", model.predict(np.array([[6]]))[0])

Rust

use linfa::prelude::*;
use linfa_linear::LinearRegression;
use ndarray::array;

fn main() {
    // dataset
    let x = array![[1.0], [2.0], [3.0], [4.0], [5.0]];
    let y = array![2.2, 4.1, 6.0, 8.1, 9.9];

    let dataset = Dataset::new(x, y);
    let model = LinearRegression::default().fit(&dataset).unwrap();

    let new_x = array![[6.0]];
    let prediction = model.predict(&new_x);
    println!("Prediction for 6: {}", prediction[0]);
}

---

7. Real-World Case Studies

• Finance: Credit scoring with logistic regression.
• Healthcare: Cancer detection using CNNs.
• Transportation: Self-driving with deep RL.
• E-commerce: Personalized recommendations with collaborative filtering.
• NLP: GPT models for chat, summarization, translation.

---

8. Connections to Next Lessons

This introduction lays the foundation. Next topics:

• Linear Regression → First algorithm in detail.
• Logistic Regression → Classification problems.
• Decision Trees → Rule-based predictive modeling.
• Model Evaluation → Metrics to judge performance.

---

9. Further Reading

• Tom Mitchell, Machine Learning.
• Bishop, Pattern Recognition and Machine Learning.
• Goodfellow et al., Deep Learning.
• scikit-learn documentation.
• Rust linfa documentation.

---