Introduction to Machine Learning

Study Notes

Machine Learning is the science and art of programming computers to learn from data.
An alternative definition is that Machine Learning gives computers the ability to learn without explicit programming.
A further definition, focused on engineering, describes a computer program learning from experience (E) in relation to a task (T).

Traditional programming for a spam filter involves identifying patterns (e.g., words like "4U" or "amazing") and writing explicit rules to flag emails containing those patterns.
This approach creates a lengthy list of rules that are complex and hard to maintain.
Machine Learning-based spam filters automatically learn patterns in spam emails compared to regular emails ("ham").
This approach produces a shorter, easier-to-maintain program, and is typically more accurate.

Machine Learning draws on various concepts and methodologies from other fields, making it a true multi-disciplinary field.
Fields that significantly overlap with Machine Learning include statistics, mathematics, data mining, computer science, deep learning, natural language processing, artificial intelligence, and more.

Supervised or unsupervised: Systems are categorized based on whether they need human supervision during training (supervised, unsupervised, semi-supervised, Reinforcement Learning).
Batch or online: Systems can be categorized by whether they learn incrementally on the fly or in batches.
Instance-Based vs. Model-Based: Machine Learning systems may classify new data points by simply comparing them to known data points or by creating predictive models.

In supervised learning, the training data includes desired solutions (labels) that the algorithm learns from.
A common example is spam classification, where the algorithm learns to distinguish spam from legitimate emails.
Other tasks within supervised learning include classification (e.g., spam detection) and regression (e.g., predicting the price of a car).