Expand description
Matrix and tensor decomposition algorithms for dimensionality reduction
This module provides various decomposition techniques including:
- PCA: Principal Component Analysis with SVD (including Randomized SVD)
- Incremental PCA: Memory-efficient PCA for large datasets
- Kernel PCA: Non-linear dimensionality reduction using kernel methods
- ICA: Independent Component Analysis (including constrained ICA)
- NMF: Non-negative Matrix Factorization
- Factor Analysis: Statistical model for latent variables
- Dictionary Learning: Sparse coding and dictionary learning
- Tensor Decomposition: CP (CANDECOMP/PARAFAC) and Tucker decomposition
- Matrix Completion: Filling missing values using low-rank matrix completion
- CCA: Canonical Correlation Analysis for finding linear relationships between two datasets
- PLS: Partial Least Squares for regression and dimensionality reduction
- Time Series: SSA, seasonal decomposition, and trend extraction
- Signal Processing: EMD, spectral decomposition, and adaptive methods
- Image & Computer Vision: 2D-PCA, image denoising, face recognition, texture analysis
- Manifold Learning: LLE, Isomap, Laplacian Eigenmaps, t-SNE, UMAP
- Component Selection: Cross-validation, bootstrap, information criteria, parallel analysis
- Quality Metrics: Goodness-of-fit statistics, reconstruction quality, interpretability measures
- Robust Methods: Robust PCA with L1 loss, M-estimators, outlier-resistant methods
- Hardware Acceleration: SIMD optimizations, parallel processing, and mixed-precision arithmetic
- Distributed Processing: Large-scale distributed decomposition across multiple nodes/workers
- Scikit-learn Compatibility: Drop-in replacements for scikit-learn transformers with full API compatibility
- Advanced Format Support: HDF5, sparse matrices, memory-mapped files, and compressed storage
- Cache Optimization: Memory-aligned data structures, tiled algorithms, and performance analysis
- Comprehensive Validation: Input validation, parameter checking, and result quality assessment
- Modular Architecture: Pluggable algorithms, preprocessing pipelines, and extensible framework
- Constrained Decomposition: Orthogonality, non-negativity, sparsity, and smoothness constraints
- Type-Safe Decomposition: Zero-cost abstractions with compile-time dimension and rank checking
Re-exports§
pub use hardware_acceleration::AccelerationConfig;pub use hardware_acceleration::AlignedMemoryOps;pub use hardware_acceleration::MixedPrecisionOps;pub use hardware_acceleration::ParallelDecomposition;pub use hardware_acceleration::SimdMatrixOps;pub use modular_framework::AlgorithmCapabilities;pub use modular_framework::AlgorithmCapability;pub use modular_framework::AlgorithmMetadata;pub use modular_framework::AlgorithmRegistry;pub use modular_framework::ComputationalComplexity;pub use modular_framework::DecompositionAlgorithm as DecompositionAlgorithmTrait;pub use modular_framework::DecompositionComponents;pub use modular_framework::DecompositionParams;pub use modular_framework::DecompositionWorkflowBuilder;pub use modular_framework::MatrixProperty;pub use modular_framework::ParamValue;pub use modular_framework::PostprocessingStep;pub use modular_framework::PreprocessingStep;pub use modular_framework::StandardizationStep;pub use modular_framework::VarimaxRotationStep;pub use robust_methods::BreakdownPointAnalysis;pub use robust_methods::BreakdownResult;pub use robust_methods::LossFunction;pub use robust_methods::MEstimatorDecomposition;pub use robust_methods::MEstimatorResult;pub use robust_methods::RobustConfig;pub use robust_methods::RobustPCAResult;pub use sklearn_compat::CrossValidation;pub use sklearn_compat::GridSearchCV;pub use sklearn_compat::ParameterValue as SklearnParameterValue;pub use sklearn_compat::SklearnPCA;pub use sklearn_compat::SklearnPipeline;pub use sklearn_compat::SklearnTransformer;pub use type_safe::CenteringOperation;pub use type_safe::ComponentAccess;pub use type_safe::ComponentIndex;pub use type_safe::DecompositionOperation;pub use type_safe::DecompositionPipeline as TypeSafeDecompositionPipeline;pub use type_safe::DecompositionState;pub use type_safe::Dimensions;pub use type_safe::Fitted;pub use type_safe::Rank;pub use type_safe::ScalingOperation;pub use type_safe::TypeSafeDecomposition;pub use type_safe::TypeSafeMatrix;pub use type_safe::TypeSafePCA;pub use type_safe::Untrained;pub use component_selection::*;pub use constrained_decomposition::*;pub use dictionary_learning::*;pub use distributed::*;pub use error_diagnostics::*;pub use factor_analysis::*;pub use fluent_api::*;pub use ica::*;pub use image_cv::*;pub use integration::*;pub use memory_efficiency::*;pub use nmf::*;pub use online_nmf::*;pub use pca::*;pub use performance::*;pub use quality_metrics::*;pub use signal_processing::*;pub use streaming::*;pub use time_series::*;pub use validation::*;pub use visualization::*;
Modules§
- cache_
optimization - Cache-Friendly Matrix Layouts and Performance Optimizations
- component_
selection - Component Selection Methods for Decomposition
- constrained_
decomposition - Constrained Decomposition Methods
- dictionary_
learning - Dictionary Learning implementation
- distributed
- Distributed Decomposition Methods for Large-Scale Processing
- error_
diagnostics - Comprehensive Error Handling and Diagnostics for Decomposition
- factor_
analysis - Factor Analysis implementation
- fluent_
api - Fluent API for Decomposition Pipelines
- hardware_
acceleration - Hardware Acceleration for Decomposition Algorithms
- ica
- Independent Component Analysis (ICA) implementation.
- image_
cv - Image and Computer Vision Decomposition Methods
- integration
- Enhanced Integration and Interoperability Module
- memory_
efficiency - Memory efficiency improvements for decomposition algorithms
- modular_
framework - Modular Pluggable Decomposition Architecture
- nmf
- Non-negative Matrix Factorization (NMF) implementation.
- online_
nmf - Online Non-negative Matrix Factorization (Online NMF)
- pca
- Principal Component Analysis and dimensionality reduction utilities
- performance
- Performance Optimization Utilities
- quality_
metrics - Quality Metrics and Goodness-of-Fit Statistics for Decomposition
- robust_
methods - Robust Decomposition Methods
- signal_
processing - Signal Processing Framework for Decomposition Applications
- sklearn_
compat - Scikit-learn Compatibility Layer
- streaming
- Real-time and streaming decomposition algorithms
- time_
series - Time Series Decomposition methods
- type_
safe - Type-safe decomposition abstractions using Rust’s type system
- validation
- Comprehensive Validation Framework for Matrix Decomposition
- visualization
- Visualization and interpretation utilities for matrix decomposition
Macros§
- s
- Slice argument constructor.