AI-ordlista
Den kompletta ordlistan över AI
Acquisition
A strategy that determines the next evaluation point by balancing the exploitation of known promising areas and the exploration of uncertain areas.
Co-fidelity
An approach where multiple data sources of different fidelities are used simultaneously to build a more accurate and efficient surrogate model.
Surrogate-Based Optimization (SBO)
A class of optimization methods that use surrogate models to approximate expensive functions, thereby reducing the number of direct evaluations needed.
Multi-Information Source Optimization (MISO)
An optimization strategy that actively integrates and exploits multiple information sources of varying quality and cost to accelerate convergence.
Scale factor
A multiplicative parameter used to scale the predictions of a low-fidelity model to align them with those of a high-fidelity model.
Black-box optimization
An optimization problem where the internal structure of the objective function is unknown and only input/output evaluation is possible.
Noise variance
A hyperparameter of the Gaussian process that quantifies the level of uncertainty or random error in the observations of the objective function.
Correlation length
A hyperparameter of the Gaussian process that determines the distance over which function points are correlated, controlling the smoothness of the model.
Initial Sampling
The initial phase of optimization where a small set of points is evaluated (often with an experimental design) to build a first version of the surrogate model.
Experimental Design
A strategy for selecting the initial evaluation points to maximize the information gathered about the objective function with a limited number of evaluations.