# TRIC-DT Glossary of Digital Twin Technology and Research
[TOC]
## Purpose and Scope

*Welcome to the TRIC-DT Glossary, a collaborative and evolving resource dedicated to enhancing understanding and communication within the digital twin reserach community*

This glossary focuses on clarifying and harmonizing the use of technical terms, jargon, and concepts central to the research methoddology and digital twinning technology across the diverse disciplines of environment, health, and infrastructure.

We invite and encourage contributions from all members of the TRIC-DT community and the wider digital twin research field.

[Template]

### TERM
**Definition**:
**Applications Health/Environment/Infrastructure**:
**Comments**:

## External glossaries
https://www.digitaltwinconsortium.org/glossary/
https://digitaltwinhub.co.uk/glossary/

# Glossary

## Agent
**Definition**: An agent is an entity that acts or has the capacity to act. It can be an individual, a group, a software program, or any entity that performs actions in a particular environment based on its design, instructions, or natural behavior.
**Applications Health/Environment/Infrastructure**:
- Social Network Modeling: An agent represents an individual or group in a social network, interacting according to social dynamics.
- Knowledge Graphs in Computing: Refers to a software program that autonomously performs tasks, such as data retrieval or graph updates ([for example here](https://www.sciencedirect.com/science/article/pii/S0167739X23003825)).

**Comments**:

## Biomarker
**Definition**: a naturally occurring molecule, gene, or characteristic by which a particular pathological or physiological process, disease, etc. can be identified.
**Applications Health**:
**Comments**:

## Clinical data
**Definition**: any quantity that can be derived from measurements performed directly from a patient (for example, scalars, time traces, 3D geometry).
**Applications Health**:
**Comments**:

## Cross-validation
**Definition**: A method to evaluate models. The idea is to split data into a training and test set, train the model on the training set and evaluate it on the test set. This is repeated multiple times with dierent splits of the data to get a more robust estimate of the model performance.

## Dynamic Knowledge Graph
**Definition**:
**Applications Health/Environment/Infrastructure**:
**Comments**:
Is similar to a relational databse but more efficient for search. Also, in contrast to a relational database, that contains the actual data, a knowledge graph defines abstractions of the data. For example if a sensor measures some data over time (e.g. one data point every second), which which is stored in the database, the knowledge graph might only retain a summary statistic (e.g average measurement each day).

## Emulation
**Definition**: Training an ML model to replicate the outputs of a mechanistic model, i.e. building a mapping between the inputs and the outputs without modelling the physics underlying the output, as a means of accelerating run times or allowing more scenarios to be tested.
**Applications Health/Environment/Infrastructure**:
- For example at Sheffield, Emulators are used to predict structural health (stress) or responsiveness of bridges
**Comments**:

## Emulator
**Definition**: also metamodel, surrogate model. A
model that approximates a simulation, but runs faster
and more efficient. Could be any machine learning
model. [Gaussian Processes](https://en.wikipedia.org/wiki/Gaussian_process_emulator) are a popular choice.

## Experimental design
**Definition**: In surrogate / emulator modeling, the systematic process of optimizing the gathering of informative data points to train the emulator. A common technique is [Latin Hypercube Sampling](https://en.wikipedia.org/wiki/Latin_hypercube_sampling).

## History matching
**Definition**: a technique aiming at restricting the ranges of the input model parameters to values where to model outputs fall within target ranges provided by clinical/experimental data.
**Applications Health/Environment/Infrastructure**:
**Comments**:

### Hyperparameters
**Definition**:Parameters of a model that are
not learned from the data, such as the kernel type in
a Gaussian Process or the number of hidden layers in
a neural network.

## Latin Hypercube Sampling

A statistical method used for sampling multivariate data, often as part of the experimental design. It divides each dimension into equally probable intervals and then samples once from each interval. This ensures that samples are evenly distributed across the entire space, which builds a good foundation to construct an emulator.

## Parameter inference
**Definition**: to determine values for the input model parameters where the model behave in accordance with available real-world data.
**Applications Health/Environment/Infrastructure**:
- In health applications, parameter inference is used to build a functional digital twin of the patient's heart, where we find parameter combinations that allow us to replicate the observed clinical data collected from that specific patient.

**Comments**:

## Sensitivity analysis
**Definition**: a systematic way to determine how a variation in a input model parameters causes changes in a model output of interest.
**Applications Health/Environment/Infrastructure**:
- In health applications, sensitivity analysis is used to link biological processes represented by input model parameters to biomarkers of interest.

**Comments**:

## Simulator
**Definition**: a (normally computationally demanding) computational model that, given the input model parameters, provides model outputs or biomarkers of interest.
**Applications Health/Environment/Infrastructure**:
**Comments**: