Skip to content

Latest commit

 

History

History
508 lines (266 loc) · 8.16 KB

CHANGELOG.md

File metadata and controls

508 lines (266 loc) · 8.16 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.0.0] - 24.05.23

Changed

  • Initial release of fedalgo, the renamed version of gwasprs, to make this repository more general to support any federated algorithms.

Added

  • Add federated Cox PH Regression cox.py.

Migration Guide

To help users transition from gwasprs to fedalgo, follow these steps:

Package Import: Update all your import statements in the codebase.

# Old import
import gwasprs

# New import
from fedalgo import gwasprs

[0.21.0]

Fixed

  • Use the numpy to perform mean imputation, the output will be either np.ndarray or jax.Array.

Added

  • Add functions for performing federated LD pruning (ld.py).
  • Add functions for plotting the manhattan and qq plots of GWAS result (gwasplot.py).

[0.20.0]

Fixed

  • Fix the precision issue by replacing cdf with sf to calculate p-value.
  • Change jax_cpu_cores to jax_dev_count for more general usage.
  • Format coding style.

[0.19.8]

Fixed

  • Fix the issue came from orthonormalize output.
  • Fix the mean imputation issue when matrix too large.
  • Fix the array type under the regression objects.

[0.19.7]

Added

  • Add expected total steps in GWASDataIterator.

[0.19.6]

Added

  • Return singular values in orthonormalize().

[0.19.5]

Fixed

  • Clean the implementation of logistic regression.

[0.19.4]

Fixed

  • Update plink2 download url.

[0.19.3]

Fixed

  • Return BlockDiagonalMatrix while indexing with interval.

[0.19.2]

Fixed

  • Adjust the implementation of create_snp_table.

Added

  • Add get_qc_metadata for removing unnecessary bfile reading step.

[0.19.1]

Fixed

  • Add array type checking in array.py and aggregations.py.

[0.19.0]

Added

  • Support iterate array over rows or columns using ArrayIterator.

[0.18.6]

Fixed

  • Add array type checking in mask.py.

[0.18.5]

Added

  • BlockDiagonalMatrix supports initializing from list and dense matrix, fromdense, fromlist and fromindex.

Fixed

  • Replace scipy.linalg.block_diag with self-defining block_diag supporting n-dimensional arrays.

[0.18.4]

Added

  • Add tolist to BlockDiagonalMatrix.

Fixed

  • Move drop_missing_samples to kwargs instead.
  • Fix logics in iterator.py.

[0.18.3]

Added

  • Add block diagonal matrix multiplication assertion.

[0.18.2]

Added

  • Make drop_missing_samples optional in GWASData.

[0.18.1]

Fixed

  • Fix the endless iteration of SNPIterator.
  • Fix the additional step for two-layer iterations.

[0.18.0]

Added

  • Intersect supports aggregate list of list.

[0.17.0]

Added

  • Covariates filtering api follows the bfile filtering.

Fixed

  • Make qc codes easy to read.

[0.16.7]

Fixed

  • Bug fix.

[0.16.6]

Fixed

  • Make poetry to build plink2.

[0.16.5]

Fixed

  • Fix the incorrect sizes of the outputs from GWASDataIterator.

[0.16.4]

Fixed

  • Remove keeping dropped information.

[0.16.3]

Fixed

  • Unit test over python 3.12.

[0.16.2]

Fixed

  • Ensure dimension of mse and XtX are matched, otherwise align them together.
  • Resolve frequent CI error due to numerical error.

[0.16.1]

Fixed

  • Increase numerical stability for inv and InverseSolver.

[0.16.0]

Fixed

  • Change BlockedLinearRegression.sse(X, y, nobs) to BlockedLinearRegression.sse(X, y).

[0.15.8]

Fixed

  • Increase numerical stability for CholeskySolver.

[0.15.7]

Added

  • BlockDiagonalMatrix supports addition only for matrix which poses the same blockshapes.

Fixed

  • SumUp supports BlockDiagonalMatrix.

[0.15.6]

Added

  • Support BlockDiagonalMatrix.shape.

Fixed

  • CholeskySolver supports BlockDiagonalMatrix.

[0.15.5]

Added

  • Support BlockDiagonalMatrix.append(BlockDiagonalMatrix).

Fixed

  • Deep copy arrays while construsting BlockDiagonalMatrix.

[0.15.4]

Fixed

  • Improve t_stats and add mse.

[0.15.3]

Fixed

  • Adapt blocked_unnorm_autocovariance and blocked_unnorm_covariance to BlockDiagonalMatrix.

[0.15.2]

Fixed

  • BlockedLinearRegression integrates BlockDiagonalMatrix.

[0.15.1]

Fixed

  • Fix GWASData.subset.
  • Change dependency polars to pandas.

[0.15.0]

Added

  • Add BlockDiagonalMatrix, AbstractBlockDiagonalMatrix and BlockDiagonalMatrixIterator.
  • Support indexing, shape, matrix-vector multiplication (linalg.mvmul), matrix-vector dot product (linalg.mvdot), matrix-matrix multiplication (linalg.matmul) and inverse operation (linalg.inv) for BlockDiagonalMatrix.

[0.14.0]

Added

  • Add BlockedLinearRegression.

[0.13.2]

Added

  • LinearRegression.predict support sparse X.

[0.13.1]

Added

  • LinearRegression support include_bias parameter.

[0.13.0]

Added

  • SumUp support sparse array.

[0.12.0]

Added

  • Add GWASData and GWASDataIterator.

[0.11.0]

Added

  • Add blocked_unnorm_autocovariance and blocked_unnorm_covariance.

[0.10.2]

Fixed

  • Refactor 4 interfaces for PCA algorithms.

[0.10.1]

Added

  • Migrate array functions from RAFAEL and add a series of array types and expand_to_2dim.

[0.10.0]

Added

  • Add Intersect.

[0.9.1]

Added

  • Refactor QC.
  • Add qc_stats, calculate_allele_freq_hwe, calculate_homo_het_count, write_snp_list, get_histogram and get_obs_count.
  • Deprecate cal_qc_client.

[0.9.0]

Added

  • Refactor loader.

[0.8.1]

Added

  • Add include_bias for BatchedLinearRegression.

Fixed

  • Correct dof for BatchedLinearRegression.
  • Fix fit construction for BatchedLinearRegression.

[0.8.0]

Added

  • Add SumUp aggregation.

[0.7.3]

Fixed

  • Fix batch axis for batched_unnorm_autocovariance.

[0.7.2]

Added

  • Add gwasprs.regression.add_bias.

[0.7.1]

Fixed

  • Match the requirement of logistic usecase in stepfl

[0.7.0]

Added

  • Add batched logistic regression

Fixed

  • Set batch axis to 0 for all batch operations.

[0.6.2]

Added

  • Extract 4 interfaces for PCA algorithms
  • Add comparisons between plink, numpy and fed-svd

[0.6.1]

Fixed

  • Match the requirement of PCA usecase in stepfl

[0.6.0]

Added

  • Main update of all federated SVD process

[0.5.4]

Fixed

  • Fix plink2 not found completely and support different platform plink2.

[0.5.3]

Fixed

  • Fix plink2 not found.

[0.5.2]

Fixed

  • swap in loader, remove self before BIM.
  • add --rm-dup 'force-first' in qc stat calculation to deal with duplication problem
  • chromosome name is set to 1..22 instead of 1..21
  • Some minor text adjustment
  • (Batched)InverseSolver solve linear system bypassing inverse matrix

[0.5.1]

Fixed

  • Switch nose to pytest.
  • Add setup_plink2 and collect to setup.py.

[0.5.0]

Added

  • batched_unnorm_autocovariance and BatchedLinearRegression support pmap.

[0.4.1]

Fixed

  • Fix BatchedCholeskySolver.

[0.4.0]

Added

  • Add batched_diagonal to linalg.

Fixed

  • Fix BatchedLinearRegression.t_stats.

[0.3.1]

Fixed

  • Fix BatchedLinearRegression.predict.

[0.3.0]

Added

  • Add batched_matmul, batched_mvdot, batched_mvdot, batched_mvmul, batched_inv, batched_cholesky, batched_solve_triangular, BatchedInverseSolver, BatchedCholeskySolver to linalg.
  • Add batched_unnorm_autocovariance, batched_unnorm_covariance to stats.
  • Add BatchedLinearRegression to regression.
  • Add isnonnan to mask.

[0.2.1]

Added

  • Add test data for test_qc and test_loader.
  • Add get_snp_table in GwasDataLoader to deal with duplicated snp.

Fixed

  • Qc, non-autosome snp will cause abnormal .hardy file (haploid will be removed, x saved as .hardy.x), implementation for filtering out non-autosome snp in loader
  • Fix snp rename max allele size to 23