v6.0.0

3D estimator is here!

What's Changed

• Continuum marginalization matrices by @p-slash in #94
• Fftlog by @p-slash in #95
• Use float in coordinates. Log2 in cf interpolation with fastlog2 bit-hacking by @p-slash in #96
• bicubic spline by @p-slash in #98
• fix bicubic spline initial point by @p-slash in #99
• Bootstrap improvements by @p-slash in #100
• Undepracate iterations by @p-slash in #101
• Precondition by @p-slash in #102
• Block diagonal small-scale randoms and perturbation by @p-slash in #103
• Apodizing function & positive definite C by @p-slash in #104
• Max eigenvalue by @p-slash in #105
• cubic hermite spline by @p-slash in #106
• Accurate calculation of small scale distance by @p-slash in #108
• bicubic spline by @p-slash in #97
• Unroll total bias by @p-slash in #107
• Estimate multipoles by @p-slash in #109
• RA & Dec cut by config by @p-slash in #110
• New square root by @p-slash in #112
• Higher order correlations by @p-slash in #111
• Si oscillations by @p-slash in #113
• Static OpenMP scheduling when marginalization by @p-slash in #114
• eliminate fpos from reading continuum marg file. by @p-slash in #115
• Cmake compilation by @p-slash in #116
• interface library for all objects by @p-slash in #117
• Save neighbors to file by @p-slash in #118
• Update fisher deriv est by @p-slash in #119
• new metal lines and their cross correlations are implemented by @p-slash in #120
• Fix MC square root by @p-slash in #121
• MC: mesh + sqrt(small-scale) by @p-slash in #122
• CMake MPI-OpenMP fix and Multipole Interpolation by @p-slash in #125
• Pade and Schulz algorithms for square root by @p-slash in #124
• estimateFrobeniusNormAs method by @p-slash in #126
• 3D optimal estimator by @p-slash in #128

Full Changelog: v5.1.1...v6.0.0

v5.1.1

What's Changed

• cubic spline with tabulated not-a-knot boundary condition by @p-slash in #89
• Small fix to redshift boundary cut by @p-slash in #90
• Redesign use N^-1 instead of N for numerical precision by @p-slash in #91
• Redesign bias estimation and limit number of iterations = 1 by @p-slash in #92
• Smooth variance instead of ivar. Option for redshift growth by @p-slash in #93

Full Changelog: v5.1.0...v5.1.1

v5.1.0

This release fixes the crucial bug in xQMLE, and adds various features.

What's Changed

• Configurable xQml estimator for different fibers by @p-slash in #77
• xQMLE with different petals. Overlap 60% by @p-slash in #78
• MPI manager by @p-slash in #79
• Smooth time in progress. Print total excepted time by @p-slash in #80
• DEBUG flag to check exposure combinations by @p-slash in #81
• MinimumSnrCut to skip forests with low SNR by @p-slash in #82
• Check outliers XE in debug by @p-slash in #83
• Critical fix: copy expid in qso_file copy constructor by @p-slash in #84
• small changes: start index 1 for printed chunk indices by @p-slash in #85
• Cubic spline by @p-slash in #86
• mpi_manager.hpp: include vector and algorithm by @p-slash in #88
• Crucial fix to xQMLE by @p-slash in #87

Full Changelog: v5.0.4...v5.1.0

v5.0.0

The cross-exposure estimator is here!

What's Changed

• Save bootstrap realizations by @p-slash in #64
• Exact expression for Q matrices by @p-slash in #65
• Global memory by @p-slash in #66
• Bootstrap improvements by @p-slash in #67
• More tempmatrices by @p-slash in #68
• round to 1 decimals in findSpecResIndex by @p-slash in #69
• Create testing.yml by @p-slash in #70
• Cross exposure by @p-slash in #71
• Outlier masking by @p-slash in #72

Full Changelog: v4.5.6...v5.0.0

v4.5.6

What's Changed

• Fix seg fault by @p-slash in #58
• Fix if mask is empty in smoothing by @p-slash in #59
• Optimize OMP and oversampled sandwich by @p-slash in #60
• new solver_invfisher_matrix variable to save inverse fisher solver, S… by @p-slash in #61
• Fix bootstrap for non-MPI case. by @p-slash in #62
• improve bootstrap speed in median by @p-slash in #63

Full Changelog: v4.5.2...v4.5.6

v4.5.2

What's Changed

• reading oversampled resolution matrix no longer supported by @p-slash in #49
• SmoothResolutionMatrix option by @p-slash in #50
• New fisher optimization with 11 diagonals by @p-slash in #51
• use tophat when nz == 1 by @p-slash in #52
• Cont. marg. improved svd condition by @p-slash in #53
• use dsymv since it is faster by @p-slash in #54
• fix median covariance by @p-slash in #55
• fix Chunk::getComputeTimeEst depracated use of preproc defines by @p-slash in #56
• _getMaxKindex applies 0.85 to have correct nkbins by @p-slash in #57

Full Changelog: v4.5.1...v4.5.2

v4.5.1

What's Changed

• Median statistics bootstrap by @p-slash in #48

Full Changelog: v4.5.0...v4.5.1

v4.5.0

What's Changed

• Cholesky in bootstrap. redshift binning function is config parameter,… by @p-slash in #38
• simplify config reading in globals by @p-slash in #39
• unique pointers in sq table by @p-slash in #40
• simplify config. common double epsilon by @p-slash in #41
• Mathtools improvements by @p-slash in #42
• Bug fix Fisher loop by @p-slash in #43
• damping based on singular value ratio jump. by @p-slash in #44
• Chunks are collapsed into OneQSOEstimate for bootstrapping by @p-slash in #45
• Improved memory tracking by @p-slash in #46
• Fast bootstrap by @p-slash in #47

Full Changelog: v4.4.4...v4.5.0

v4.4.0

What's Changed

The main algorithmic change is that weighted Q matrices are now C-1 Q. This needed fast transpose copy. Furthermore, time cost estimation function was updated to include resolution matrix multiplications and correct number of weighted Q matrices. There are also attempts at optimizing setQ function.

• vectorized velocity and redshift interpolation functions by @p-slash in #23
• compute C-1 Qi instead of C-1 Qi C-1. by @p-slash in #24
• Combine index and stored Q matrices and vectorize by @p-slash in #25
• eliminate temp z array and vectorized interpolation by @p-slash in #26

Full Changelog: v4.3.6...v4.4.0

v4.3.6

What's Changed

• Test bootstrap timings by @p-slash in #21

Full Changelog: v4.3.5...v4.3.6