Basic Concepts

1. What is Machine Learning?#

Machine learning (ML) is a field of artificial intelligence that focuses on building systems that can learn patterns from data and improve their performance on a task without being explicitly programmed.

Instead of relying on hard-coded rules, ML systems use algorithms to map inputs (features) to outputs (predictions or decisions).

At its core, ML answers the question:

“Given data, how can we design models that improve through experience?”

2. Common Learning Methods#

Machine learning can be categorized in different ways. Two common perspectives are based on the learning target and based on the learning strategy.

2.1 Based on Learning Target#

• Supervised Learning
  Learn from labeled data (input–output pairs).
  Examples: image classification, house price prediction.

• Unsupervised Learning
  Learn from unlabeled data by finding hidden structures.
  Examples: clustering customers, reducing dimensions (PCA).

• Self-supervised Learning
  Learn supervisory signals from the data itself (a hybrid of supervised + unsupervised).
  Examples: predicting masked words in a sentence (BERT), contrastive learning.

• Reinforcement Learning (RL)
  Learn by interacting with an environment, receiving feedback as rewards or penalties.
  Examples: game-playing agents, robotic control.

2.2 Based on Learning Strategy#

• Parametric vs. Non-parametric Models

  • Parametric: fixed number of parameters regardless of dataset size (e.g., linear regression, neural networks).
  • Non-parametric: model complexity grows with data (e.g., k-Nearest Neighbors, decision trees).

• Generative vs. Discriminative Models

  • Generative: model how data is generated (e.g., Naive Bayes, Variational Autoencoders).
  • Discriminative: focus only on decision boundaries between classes (e.g., logistic regression, SVM).

• Online vs. Batch Learning

  • Batch: train once on the whole dataset.
  • Online: update incrementally as new data arrives.

3. Key Components of Machine Learning#

These are the core pillars of ML, which are also fundamental for deep learning. Each links to a dedicated section:

• Data — the foundation; includes features, labels, preprocessing, and data quality.
• Model — the function or algorithm that learns patterns from data.
• Training — the process of fitting the model to data, often involving optimization.
• Evaluation — assessing how well the model generalizes using appropriate metrics.

💡 Together, these sections set the stage for deeper dives into data, models, training, and evaluation in the upcoming chapters.