obj = createns (X) creates a nearest neighbor searcher object using the training data X. By default, it constructs an ExhaustiveSearcher object with the Euclidean distance metric.

obj = createns (X, name, value, …) allows customization of the searcher type and its properties through name-value pairs. The following name-value pair is supported to specify the searcher type:

Additional name-value pairs depend on the selected 'NSMethod' and are passed directly to the constructor of the corresponding class:

• For 'exhaustive', see ExhaustiveSearcher documentation for parameters like 'Distance', 'P', 'Scale', and 'Cov'.
• For 'kdtree', see KDTreeSearcher documentation for parameters like 'Distance', 'P', and 'BucketSize'.
• For 'hnsw', see hnswSearcher documentation for parameters like 'Distance', 'P', 'Scale', 'Cov', 'MaxNumLinksPerNode', and 'TrainSetSize'.

Input Arguments:

• X - Training data, specified as an $N\times P$ numeric matrix where rows represent observations and columns represent features. Must be finite and numeric.

Output:

• obj - A nearest neighbor searcher object of type ExhaustiveSearcher, KDTreeSearcher, or hnswSearcher, depending on the specified 'NSMethod'.

Examples:

 ## Create an ExhaustiveSearcher with default parameters
 X = [1, 2; 3, 4; 5, 6];
 obj = createns (X);

 ## Create a KDTreeSearcher with Euclidean distance
 obj = createns (X, "NSMethod", "kdtree", "Distance", "euclidean");

 ## Create an hnswSearcher with Minkowski distance and custom parameters
 obj = createns (X, "NSMethod", "hnsw", "Distance", "minkowski", "P", 3, ...
                 "MaxNumLinksPerNode", 2);

See also: ExhaustiveSearcher, KDTreeSearcher, hnswSearcher, knnsearch, rangesearch

Source Code: createns