AI-ordlista
Den kompletta ordlistan över AI
Density Mode
Point in the feature space where the data's probability density reaches a local maximum, to which the Mean Shift algorithm converges to identify the center of a cluster.
Search Window
Circular or spherical region around a data point, defined by the bandwidth, within which the algorithm calculates the weighted average of points to determine the direction of movement.
Displacement Vector
Vector calculated at each iteration of the algorithm, pointing from a point's current position towards the weighted center of mass of its neighbors, thus guiding its progression towards the density mode.
Mean Shift Trajectory
Path traveled by a data point through successive iterations of the algorithm, from its initial position to its final convergence position on a density mode.
Image Segmentation
Flagship application of the Mean Shift algorithm, where pixels are treated as points in a feature space (color, position) to group homogeneous regions of the image.
Combined Feature Space
Multidimensional space used for image segmentation, combining spatial attributes (x, y coordinates) and appearance attributes (RGB or Lab color values) to group pixels.
Kernel Density Estimation (KDE)
Statistical method on which Mean Shift is based, which estimates the probability density function of a dataset by summing kernel functions centered on each point.
Grid Search
Optimization technique to find the best bandwidth by evaluating clustering performance over a predefined range of values, often based on metrics like the silhouette score.
Mode Seeking
Alternative name and fundamental concept of Mean Shift, describing the nature of the algorithm as a procedure for finding the maxima (modes) of the underlying data density function.
Anchor Point
In some implementations, a subset of data points used to initialize the algorithm to reduce computational complexity, with other points then assigned to the nearest mode.