AI-ordlista
Den kompletta ordlistan över AI
Combinatorial Bayesian Optimization
Adaptation of Bayesian optimization to discrete or combinatorial search spaces, using specific surrogate models to handle structures like graphs or permutations.
Categorical Surrogate Model
Surrogate model designed to handle categorical or discrete variables, often based on Gaussian processes with kernels adapted to discrete spaces.
Hamming Kernel
Specific kernel function for discrete spaces that measures similarity between two points by counting the number of different coordinates, commonly used in Gaussian processes for combinatorial optimization.
Lattice Acquisition
Acquisition strategy that explores the discrete search space by following a lattice structure, allowing systematic evaluation of neighboring configurations.
BOCP (Bayesian Optimization for Combinatorial Problems)
Specific methodological framework for applying Bayesian optimization to combinatorial problems, integrating adapted models and acquisition strategies.
Permutation Space
Discrete search domain where solutions are ordered arrangements of elements, requiring specialized similarity metrics and kernels like the Kendall kernel.
Graph Kernel
Kernel function defined on graph structures that computes similarity between two graphs based on their topological properties or common substructures.
Random Markov Model
Alternative approach to Gaussian process for modeling the objective function in discrete spaces, capturing dependencies between binary or categorical variables.
Multi-Objective Combinatorial Optimization
Extension of combinatorial Bayesian optimization to problems with multiple conflicting objectives, using approximate Pareto frontiers in discrete spaces.
One-Hot Representation
Encoding technique for categorical variables into binary vectors to enable the use of continuous models in combinatorial optimization contexts.
Partition Tree Method
Approach that recursively divides the discrete search space into sub-regions using decision trees, guided by objective function observations.
BO with Mixed Variables
Variant of Bayesian optimization simultaneously handling continuous, discrete, and categorical variables, requiring hybrid surrogate models.
Simulated Annealing Acquisition
Acquisition strategy that combines Bayesian criteria with a simulated annealing mechanism to escape local optima in discrete landscapes.
Tree-Based Surrogate Model
Alternative to Gaussian processes using ensemble models like random forests, naturally suited for discrete spaces and non-linear structures.
Kendall Distance
Similarity metric between permutations that counts the minimum number of adjacent swaps needed to transform one permutation into another, used in kernels for ordering spaces.
Sequential Bayesian Optimization
Application of Bayesian optimization to sequential decision problems where actions are discrete, modeling the optimal policy with Gaussian processes.
String Kernel
Specialized kernel function for string spaces or discrete sequences, computing similarity based on common subsequences.
BO for Discrete Hyperparameters
Specific application of combinatorial Bayesian optimization for hyperparameter tuning when these belong to discrete or categorical sets.