Machine Learning: Supervised, Unsupervised, and Reinforcement Learning

This set of vocabulary flashcards covers the fundamentals of Machine Learning, including its definitions, the three primary paradigms (Supervised, Unsupervised, Reinforcement Learning), and specific tasks and models as presented by Prof. Dr. Saad Albawi.

Learning

To gain knowledge or understanding of, or skill in by study, instruction or experience, which includes learning new facts, learning how to do something, or improving an existing ability.

Machine Learning (ML)

The study of methods for programming computers to learn automatically from experience, typically involving tasks like recognition, diagnosis, planning, robot control, and prediction.

Training dataset

The initial data provided to a machine to provide experience, allowing it to identify rules, patterns, and relations to derive inferences.

Supervised Learning

A machine learning paradigm where the machine is provided with a labelled training dataset (Samples + Labels) to devise a function that produces an output value for new samples.

Classification

A type of supervised learning where the possible output values of the function are predefined and discrete or categorical, such as identifying if an object is a CAR or a BIKE.

Regression

A type of supervised learning used when the possible output values of the function are continuous real values, such as predicting a price of $20500.50$ in the year $2025$.

Unsupervised Learning

A paradigm where the learner is provided only with unlabeled data and no feedback from the environment; it identifies statistical structures and patterns within the input collection.

Clustering

An unsupervised learning task that groups similar objects together into homogeneous clusters that are not predefined, such as market segmentation or anomaly detection.

Association Rules Mining

A task in unsupervised learning used to identify relations between items, commonly seen in examples like 'Customers who viewed this item also viewed…'

Reinforcement Learning

Also known as learning from trials and errors, where an agent learns tasks by interacting with an environment through actions, rewards, feedbacks, and punishments.

Agent

The entity in reinforcement learning that performs actions within an environment and receives rewards or punishments to learn a specific task.

Semi-Supervised Learning

A technique between supervised and unsupervised learning that utilizes a small amount of labeled data and a large amount of unlabeled data to reduce required supervision while improving results.

Six Steps in Machine Learning

$1$) Data collection, $2$) Representation, $3$) Modeling, $4$) Estimation, $5$) Validation, and $6$) Apply learned model to new 'test' data.

Bayesian learning

A specific type of machine learning model mentioned alongside Logistic Regression and Decision Trees as examples for Classification and Regression.

K-Nearest neighbor

A specific machine learning model used for Classification and Regression tasks.

DBSCAN

An example of a specific clustering model mentioned in the lecture alongside K-means and Hierarchical clustering.

Support Vector Machine

A classification and regression model listed among machine learning examples like Neural Networks and Deep Learning.

Boosting

A category of machine learning models that includes Random Forests and Xgboost.

Computational complexity

One of the listed disadvantages of Machine Learning along with noisy sensors and application-specific algorithms.

Ground Truth

The labels or values present in a dataset that the decision in supervised learning depends on, representing the documented 'experience'.