Model

A model is a learned or designed representation that maps inputs to outputs for a defined task or family of tasks.

Plan for this page

Understand one concept in plain language and connect it to nearby ideas.

You will leave with: In machine learning, training selects parameter values that shape the mapping. The model is only one layer of a working system, which may also include instructions, retrieval, tools, user interfaces, monitoring, and human decisions.
Time: 3–8 min
Before you begin: No prerequisite. Keep one real example in mind.
Do this now: Read the explanation, then compare the example with the common misunderstanding.

Prepared next stepSeparate model, product, and system

ModelA model is a learned or designed representation that maps inputs to outputs for a defined task or family of tasks.

Parameters
Training
Inference
Evaluation

Plain explanation

In machine learning, training selects parameter values that shape the mapping. The model is only one layer of a working system, which may also include instructions, retrieval, tools, user interfaces, monitoring, and human decisions.

Example

A weather model can receive measurements and produce a forecast, while a language model receives token context and predicts possible continuations.

Why it matters

Separating model from product prevents mistaken conclusions about what was tested and what can fail. A capable model can be used inside a poorly controlled system, and a modest model can be useful in a well-designed process.

Common misunderstanding

Misconception: the model is the entire AI application. Correction: applications wrap models with many technical and organizational components.

Related concepts

Follow relationships, not only the alphabet

ParametersParameters are internal numerical values that a model’s training procedure adjusts to shape its behavior.TrainingTraining is the process of adjusting a model to improve performance on an objective using data, feedback, or interaction.InferenceInference is running a trained model on an input to produce a prediction, generation, score, or other output.EvaluationEvaluation is the structured measurement and investigation of how a model or system performs against defined questions, criteria, and contexts.

Separate model, product, and system

Sources and scope

What supports this object

Machine Learning GlossaryGoogle for Developers

Technical model vocabulary for machine learning; a deployed product usually includes additional data, software, interfaces, policies and people.

Open source

Loading page content...

…