Module kernel_utils 

Kernel utility functions for machine learning workflows.

This module provides practical utilities for kernel-based machine learning:

• Kernel-target alignment for measuring kernel quality
• Gram matrix operations (eigendecomposition preparation)
• Distance matrix computation from kernels
• Kernel matrix validation

Functions

compute_gram_matrix

Compute the kernel matrix from data using a given kernel function.

distances_from_kernel

Compute pairwise distances from a kernel matrix.

estimate_kernel_rank

Compute the effective dimensionality (rank) of a kernel matrix based on normalized eigenvalue spectrum.

is_valid_kernel_matrix

Check if a kernel matrix is valid (symmetric and positive semi-definite).

kernel_target_alignment

Compute kernel-target alignment (KTA) between a kernel matrix and target labels.

median_heuristic_bandwidth

Compute kernel bandwidth using median heuristic.

normalize_rows

Normalize each row of a data matrix (L2 normalization).