Module kernel_framework 

Comprehensive trait-based framework for kernel approximations

This module provides a unified trait system for implementing kernel approximation methods, making it easy to create new approximation strategies and compose them.

Architecture

• KernelMethod: Core trait for all kernel approximation methods
• SamplingStrategy: Abstract sampling strategies (uniform, importance, etc.)
• FeatureMap: Abstract feature transformations
• ApproximationQuality: Quality metrics and guarantees
• ComposableKernel: Combine multiple kernels

Structs

CompositeKernelMethod

Composable kernel that combines multiple kernel methods

ErrorBound

Error bound information

KMeansSampling

K-means based sampling strategy

KernelAlignmentMetric

Kernel alignment quality metric

UniformSampling

Uniform random sampling strategy

Enums

BoundType

Type of error bound

CombinationStrategy

Strategy for combining multiple kernels

Complexity

Computational complexity classification

KernelType

Kernel type classification

Traits

ApproximationQuality

Approximation quality metrics

FeatureMap

Feature map transformation

KernelMethod

Core trait for kernel approximation methods

SamplingStrategy

Sampling strategy for selecting landmarks/components