CR-Sparse v0.4.0

This is a maintenance release to align the library with JAX 0.4.x versions.

CR-Sparse v0.3.2

Documentation

Added

Dictionaries

• Grassmannian frames

Linear operators

• windowed_op

Sparse Recovery algorithms

• FOCUSS
• SPGL1 (Spectral Projected Gradient L1)

Optimization : Smooth functions

• smooth_quad_error

Optimization algorithms

• Spectral projected gradient

Test Problems

• New test problems module introduced
• heavi-sine:fourier:heavi-side
• blocks:haar
• cosine-spikes:dirac-dct
• complex:sinusoid-spikes:dirac-fourier
• cosine-spikes:dirac-dct:gaussian
• piecewise-cubic-poly:daubechies:gaussian
• signed-spikes:dirac:gaussian
• complex:signed-spikes:dirac:gaussian
• blocks:heavi-side
• blocks:normalized-heavi-side
• gaussian-spikes:dirac:gaussian
• src-sep-1

Examples

• Matching pursuit demo
• Grassmannian frames demo
• Several examples based on the test problems

Documentation

• Thinking in JAX tutorial added
• Quick start expanded
• Test problems documentation linked with examples

Changed

• Matching pursuit implementation revamped

Fixed

• Handling of complex signals in Subspace Pursuit
• Handling of complex signals in Compressive Sampling Matching Pursuit

Improved

• Support change condition added in convergence criteria for Subspace Pursuit
• order attribute in reshape linear operator

Removed

• cr.sparse.io moved to cr-nimble project

CR-Sparse v0.3.1

The main highlight of this release is the support for block sparse bayesian learning algorithms.

Added

Block Sparse Bayesian Learning

• Expectation Maximization version
• Bound Optimization version

Dictionaries

• sparse_binary_mtx
• wavelet_basis

Linear operators

• sparse_real_matrix
• sparse_binary_dict

Test data

• sparse_normal_blocks

Miscellaneous

• Some plotting utilities

Examples

• ECG Data Compressive Sensing
• Block Sparse Bayesian Learning
• Sparse Binary Sensing Matrices

Improvements

• Fixed some issues in the circulant linear operator
• Removed unnecessary __init__ files
• Removed bio module which was empty
• Resolved some warnings related to incorrect static argument names or numbers
• Added __str__ to several named tuples for debugging purposes
• ADMM tutorial updated to align with 0.3.x changes
• Added length and x properties in RecoverySolution

Others

• Changed copyright to CR-Suite Development Team

CR-Sparse v0.3.0

Added

Indicators

• zero
• singleton
• affine
• box
• box affine
• conic
• l2 ball
• l1 ball

Projectors

• zero
• identity
• singleton
• affine
• box
• conic
• l2 ball
• l1 ball

Linear operators

Proximal operators

• zero
• 1l
• l2
• l1 positive
• l1 ball
• ordered weighted l1

Smooth functions

• constant
• entropy
• huber
• linear
• logdet
• quad matrix

First order methods

• l1 regulated least square
• smooth conic dual solver
• lasso
• ordered weighted l1 regularized least squares
• dantzig smooth conic dual
• basis pursuit smooth conic dual

Removed

• Common utility functions, linear algebra routines
  and basic signal processing functions have been refactored into a separate
  library CR-Nimble
• Wavelets related functionality has been refactored into a separate
  library CR-Wavelets

Misc

• Support for JAX 0.3.14

CR-Sparse v0.2.2

Improved

• Documentation
  • Introduction page revised
  • API docs improved
  • README revised
  • Algorithm page added
  • Quick start page revised

Added

• JOSS Paper added and revised based on feedback from reviewers
• Linear Operators
  • input_shape, output_shape attributes introduced
  • fft, dot, norm estimate, reshape, scalar_mult, total variation,

CR-Sparse v0.2.1

Documentation

Added

• Compressive Sensing
  • 1-bit compressive sensing process
  • BIHT (Binary Iterative Hard Thresholding) algorithm for signal reconstruction from 1-bit measurements

CR-Sparse v0.2.0

Added

• Linear Operators
  • Convolution 1D, 2D, ND
  • Gram and Frame operators for a given linear operator
  • DWT 1D operator
  • DWT 2D operator
  • Block diagonal operator (by combining one or more operators)
• Sparse Linear Systems
  • Power iterations for computing the largest eigen value of a symmetric linear operator
  • LSQR solver for least squares problems with support for N-D data
  • ISTA: Iterative Shrinkage and Thresholding Algorithm
  • FISTA: Fast Iterative Shrinkage and Thresholding Algorithm
  • lanbpro, simple lansvd
• Geophysics
  • Ricker wavelet
  • Hard, soft and half thresholding operators for ND arrays (both absolute and percentile thresholds)
• Image Processing
  • Gaussian kernels
• Examples
  • Deconvolution
  • Image Deblurring
• Data generation
  • Random subspaces, uniform points on subspaces
  • two_subspaces_at_angle, three_subspaces_at_angle
  • multiple index_sets
  • sparse signals with bi-uniform non-zero values
• Utilities
  • More functions for ND-arrays
  • Off diagonal elements in a matrix, min,max, mean
  • set_diagonal, abs_max_idx_cw, abs_max_idx_rw
• Linear Algebra
  • orth, row_space, null_space, left_null_space, effective_rank
  • subspaces: principal angles, is_in_subspace, project_to_subspace
  • mult_with_submatrix, solve_on_submatrix
  • lanbpro, simple lansvd
• Clustering
  • K-means clustering
  • Spectral clustering
  • Clustering error metrics
• Subspace clustering
  • OMP for sparse subspace clustering
  • Subspace preservation ratio metrics

A paper is being prepared for JOSS.

Improved

• Linear Operators
  • Ability to apply a 1D linear operator along a specific axis of input data
  • axis parameter added to various compressive sensing operators
• Code coverage
  • It is back to 90+% in the unit tests

CR.Sparse v0.1.6

Added

Wavelets

• CWT implementation based on PyWavelets: CMOR and MEXH
• integrate_wavelet, central_frequency, scale2frequency

Examples

• CoSaMP step by step
• Chirp CWT with Mexican Hat Wavelet
• Frequency Change Detection using DWT
• Cameraman Wavelet Decomposition

Changed

Wavelets

• CWT API has been revised a bit.

Updated

Examples

• Sparse recovery via ADMM

Signal Processing

• frequency_spectrum, power_spectrum

CR.Sparse v0.1.5

Linear Operators
- Orthogonal basis operators: Cosine, Walsh Hadamard
- General Operators: FIR, Circulant, First Derivative, Second Derivative, Running average
- Operators: Partial Op
- DOT TEST for linear operators added.

Convex optimization algorithms
- Sparsifying basis support in yall1
- TNIPM (Truncated Newton Interior Points Method) implemented.

Wavelets
- Forward DWT
- Inverse DWT
- Padding modes: symmetric, reflect, constant, zero, periodic, periodization
- Wavelet families: HAAR, DB, MEYER, SYMMETRIC, COIFLET
- Full DWT/IDWT for periodization mode
- Filtering with Upsampling/Downsampling
- Quadrature Mirror Filters
- Forward and inverse DWT along a specific axis
- 2D Forward and inverse DWT for images
- Print wavelet info
- wavelist
- families
- build_wavelet
- wavefun
- CWT for Morlet and Ricker wavelets introduced

Benchmarking
- Introduced airspeed velocity based benchmarks

General stuff
- Examples gallery introduced
- Unit test coverage is now back to 90%
- Documentation has been setup at ReadTheDocs.org also

CR.Sparse v0.1.4

Major updates

• A framework for linear operators
• ADMM based algorithms for l1 minimization
• Several greedy algorithms updated to support linear operators as well as plain matrices
• Hard thresholding pursuit added