Sampling
Selecting a subset of data from a larger dataset to make statistical inferences or train models.
Common AI terms beginning with S, defined for advertising professionals.
Find the Flux+Form AI glossary & dictionary to help you make sense of common AI terms. Below you can find an AI Glossary & Dictionary for “S”.
Selecting a subset of data from a larger dataset to make statistical inferences or train models.
Adjusting numerical data into a specified range to improve learning stability.
In optimization, a plan for adjusting the learning rate over the course of training.
Making a brand findable wherever people search, across AI assistants, social, marketplaces, and maps, not just Google.
A mechanism that allows models to weigh the importance of different parts of an input sequence relative to each other.
Learning from data without explicit labels by creating proxy tasks, such as predicting the next word in a sentence.
Semantic degradation is the gradual loss of specific meaning and nuance when text passes through AI again and again, leaving copy that reads smoothly but says less and less.
Assigning a class label to each pixel in an image to delineate objects and backgrounds.
Using NLP techniques to determine the emotional tone of text, such as positive, negative or neutral.
Models designed to handle sequential data, such as time series or language, by capturing order and context.
Data where order matters, such as clickstreams, sensor readings or sentences.
AI tools employees use for work without their company’s approval, oversight, or knowledge.
How often, and how favorably, a brand gets mentioned inside AI-generated answers, compared to how it ranks on a traditional search page.
An activation function that maps input values into a range between 0 and 1.
A measure that quantifies how alike two data points are, used in clustering and recommendation.
A matrix factorisation technique used for dimensionality reduction and latent semantic analysis.
A direct connection that bypasses one or more layers in a neural network, helping to combat vanishing gradients.
A compact AI language model built to run cheaper and faster than a frontier LLM on a narrower set of tasks.
An activation function that converts a vector of values into probabilities that sum to 1.
Building with AI from a clear written specification first, so the output can be checked against a stable target.
Purpose-built AI models designed for specific industries, domains, or tasks rather than general-purpose systems, delivering superior performance at lower cost than generalist models.
The proportion of true negatives correctly identified by a model, important for evaluating classifiers.
Converting spoken language into text using AI models and signal processing.
Sponsored Agents are AI tools that build and manage paid ads inside a platform, connected to the advertiser’s own commerce and CRM data.
Transforming data to have zero mean and unit variance.
In reinforcement learning, the current situation that the agent observes and uses to decide actions.
Methods that use statistics to infer relationships from data and make predictions.
An iterative optimisation algorithm that updates model parameters using random subsets (batches) of data.
Recombining the style of one image with the content of another using neural networks, useful for creative campaigns.
A narrowly scoped AI agent that a main agent hands one specific job to, with its own instructions and its own clean context.
Training models on labelled data where the correct output is provided for each input.
A supervised learning algorithm that finds the optimal hyperplane separating classes in the feature space.
When an AI model tells you what you want to hear instead of what is true.
AI generated personas that stand in for real customers, used to pre-test messaging and creative before live research or media.
Data artificially generated to augment training sets or protect privacy while preserving statistical properties.
Fully AI-generated personas, with faces, voices, and personalities, built to create content and endorse products the way a human influencer would.
The consumer reflex to read polished content as possibly AI-generated, which inverts decades of production value as a trust signal.
A collection of AI agents, each handling a defined function, working alongside human staff as a team rather than a tool.
The dictionary defines the words. The Creative Cadence Workshop builds the practice. Eight weeks of hands-on AI training built for ad agencies and in-house creative teams.
Learn about the workshop →