Learn the about the technology behind A.I.

Machine Learning Vocabulary

A set of rules or mathematical instructions given to an AI system to help it learn from data. It is the "recipe" before the cooking begins.

The specific representation learned by an algorithm after training on a dataset. If the algorithm is the recipe, the model is the "dish" ready to be served (used).

The initial dataset used to teach the machine learning model. The quality and quantity of this data directly determine the model's success.

A type of ML where the model learns from labeled data (input-output pairs). It’s like a teacher providing the questions and the correct answers.

A type of ML where the model looks for patterns in unlabeled data. It’s like asking the computer to "sort these objects" without telling it what the categories are.

A type of ML where an agent learns to make decisions by performing actions and receiving rewards or penalties (trial and error).

A subset of ML based on artificial neural networks with multiple layers (hence "deep"). It drives advanced capabilities like image recognition and NLP.

A computing system inspired by the biological neural networks of animal brains, designed to recognize patterns and interpret sensory data.

A failure where a model learns the training data too well (memorizing noise), making it perform poorly on new, unseen data.

Systematic and repeatable errors in a computer system that create unfair outcomes, often stemming from prejudiced training data.

An individual measurable property or characteristic of a phenomenon being observed. In a spreadsheet, these are the columns (e.g., "Customer Age," "Purchase History").

The "answer" or result portion of a training set. For a spam detector, the emails are the features, and "Spam" or "Not Spam" is the label.

The process of using a trained model to make predictions on live, real-world data. This is the "production" phase of ML.

Information provided by direct observation (empirical evidence) as opposed to information provided by inference. It is the reality checking of the model.

A branch of AI/ML focused on the interaction between computers and humans through natural language (text and speech).

A field of AI/ML that enables computers to derive meaningful information from digital images, videos, and other visual inputs.

The use of data, statistical algorithms, and ML techniques to identify the likelihood of future outcomes based on historical data.

A complex model (like deep learning) where the internal decision-making process is not easily understood or interpreted by humans.

Techniques and methods used to make the results of ML models understandable to human experts, countering the "black box" problem.

Key performance metrics. Precision is "how many selected items are relevant?" Recall is "how many relevant items are selected?"

The human-led process of adding tags or labels to raw data (like images or text) so that a machine learning model can understand it.

A set of practices that aims to deploy and maintain ML models in production reliably and efficiently, similar to DevOps in software engineering.

A configuration that is external to the model and whose value cannot be estimated from data (e.g., the learning rate). These are the "knobs" engineers turn to tune performance.

A type of ML capable of generating new content (text, images, code) in response to prompts, rather than just analyzing existing data.

A model where human interaction is required to train, tune, or test the algorithm, ensuring higher accuracy and safety in critical decisions.