Artificial intelligence (AI) systems, in general, are engineered systems that generate outputs such as content, forecasts, recommendations or decisions for a given set of human-defined objectives. AI covers a wide range of technologies that reflect different approaches to dealing with these complex problems.
ML is a branch of AI that employs computational techniques to enable systems to learn from data or experiences. In other words, ML systems are developed through the optimisation of algorithms to fit to training data, or improve their performance based through maximizing a reward. ML methods include deep learning, which is also addressed in this document.
Terms such as knowledge, learning and decisions are used throughout the document. However, it is not the intent to anthropomorphize machine learning (ML).
1 Scope
This document establishes an Artificial Intelligence (AI) and Machine Learning (ML) framework for describing a generic AI system using ML technology. The framework describes the system components and their functions in the AI ecosystem. This document is applicable to all types and sizes of organizations, including public and private companies, government entities, and not-for-profit organizations, that are implementing or using AI systems.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content constitutes requirements of this document. For dated references, only the edition cited applies. For undated references, the latest edition of the referenced document (including any amendments) applies.
ISO/IEC 22989, Information technology— Artificial intelligence — Artificial intelligence concepts and terminology
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 22989 and the following apply.
3.1 Model development and use
3.1.1
classification model
<machine learning> machine learning model whose expected output for a given input is one or more classes
3.1.2
regression model
<machine learning> machine learning model whose expected output for a given input is a continuous variable
3.1.3
generalization
<machine learning> ability of a trained model to make correct predictions on previously unseen input data
Note 1 to entry: A machine learning model that generalizes well is one that has acceptable prediction accuracies using previously unseen input data.
Note 2 to entry: Generalization is closely related to overfitting. An overfit machine learning model will not generalize well as the model fits the training data too precisely.
3.1.4
overfitting
<machine learning> creating a model which fits the training data too precisely and fails to generalize on new data
Note 1 to entry: Overfitting can occur because the trained model has learned from non-essential features in the training data (i.e. features that do not generalize to useful outputs), excessive noise in the training data (e.g. excessive number of outliers) or because the model is too complex for the training data.
Note 2 to entry: Overfitting can be identified when there is a significant difference between errors measured on training data and on separate test and validation data. The performance of overfitted models is especially impacted when there is a significant mismatch between training data and production data.
3.1.5
underfitting
<machine learning> creating a model that does not fit the training data closely enough and produces incorrect predictions on new data
Note 1 to entry: Underfitting can occur when features are poorly selected, insufficient training time or when the model is too simple to learn from large training data due to limited model capacity (i.e. expressive power).
3.2 Tools
3.2.1
backpropagation
neural network training method that uses the error at the output layer to adjust and optimise the weights for the connections from the successive previous layers
3.2.2
learning rate
step size for a gradient method
Note 1 to entry: Learning rate determines whether and how fast a model converges to an optimal solution, making it an important hyperparameter to set for neural networks.
3.3 Data
3.3.1
class
human-defined category of elements that are part of the dataset and that share common attributes
EXAMPLE:
"telephone", "table", "chair", "ball bearing" and "tennis ball" are classes. The "table" class includes: a work table, a dining table, a study desk, a coffee table, a workbench.
Note 1 to entry: Classes are typically target variables and designated by a name.
3.3.2
cluster
automatically induced category of elements that are part of the dataset and that share common attributes
Note 1 to entry: Clusters do not necessarily have a name.
3.3.3
feature
<machine learning> measurable property of an object or event with respect to a set of characteristics
Note 1 to entry: Features play a role in training and prediction.
Note 2 to entry: Features provide a machine-readable way to describe the relevant objects. As the algorithm will not go back to the objects or events themselves, feature representations are designed to contain all useful information.
3.3.4
distance
<machine learning> measured proximity of two points in space
Note 1 to entry: Euclidean, or straight-line, distance is ordinarily used in machine learning.
3.3.5
unlabelled
property of a sample that does not include a target variable
Open only the standard information section. To view the complete content, you need to purchase standards through the ISO online browsing platform.
https://www.iso.org/obp/ui/#iso:std:iso-iec:23053:ed-1:v1:en
