Machine Learning (ML) is a subfield of Artificial Intelligence (AI) that focuses on enabling computers to learn automatically from data without being explicitly programmed.
As defined by Arthur Samuel:
“Machine learning is the field of study that gives computers the ability to learn without being explicitly programmed.”
Difference Between Traditional Programming and Machine Learning
| Traditional Programming | Machine Learning |
|---|---|
| Programmer writes explicit logic | System learns patterns from data |
| Rules + Data → Output | Data + Output → Model (learned rules) |
| Difficult for complex problems | Effective for real-world, complex, large-scale data |
| Used in deterministic systems | Used in intelligent, adaptive systems |
Limitations of Early AI Systems
- Early AI focused on expert systems (e.g., MYCIN) which required manual rule creation by experts.
- These systems lacked true intelligence as they were heavily dependent on human-defined logic.
- They couldn’t adapt or scale well to real-world complex tasks like image recognition, natural language understanding, etc.
Modern Machine Learning Approach
Modern ML is data-driven. Instead of writing logic manually:
- Data is used to train models, which then predict outcomes or classify inputs.
- These models are capable of improving over time as more data (experience) becomes available.
Tom Mitchell’s definition:
“A computer program is said to learn from experience E, with respect to some task T and performance measure P, if its performance at task T, as measured by P, improves with experience E.”
Example:
- Task (T): Detecting objects in an image
- Experience (E): Training on thousands of labeled images
- Performance Measure (P): Accuracy, precision, recall, etc.
Learning in Humans vs. Machines
| Humans | Machines |
|---|---|
| Gain experience through observation, imitation, trial-error | Gain experience through data collection and model training |
| Form abstract concepts and generalize | Perform abstraction and generalization through models |
| Use heuristics (rules of thumb) to make decisions | Use trained models and evaluation measures to decide |
Steps in Machine Learning Process (Like Human Thinking)
- Data Collection: Gather raw input data.
- Abstraction: Identify patterns and features (like recognizing an object).
- Generalization: Apply knowledge to unseen data (learning rules).
- Heuristics Formation: Build experience to guide future decisions.
- Evaluation: Measure performance and improve model (course correction).
Types of Models in ML
- Equations
- Relational Diagrams
- Logical Rules
- Clusters or Groupings
A model is a global, reusable form of knowledge, while a pattern is local and specific.