v2.2.0

• Added pylops.signalprocessing.NonStationaryConvolve3D operator
• Added nd-array capabilities to pylops.basicoperators.Identity and pylops.basicoperators.Zero
• Added second implementation in pylops.waveeqprocessing.BlendingContinuous which is more
  performant when dealing with small number of receivers
• Added forceflat property to operators with ambiguous rmatvec (pylops.basicoperators.Block,
  pylops.basicoperators.Bilinear, pylops.basicoperators.BlockDiag, pylops.basicoperators.HStack,
  pylops.basicoperators.MatrixMult, pylops.basicoperators.VStack, and pylops.basicoperators.Zero)
• Improved dynamic mode of pylops.waveeqprocessing.Kirchhoff operator
• Modified pylops.signalprocessing.Convolve1D to allow both filters that are both shorter and longer of the
  input vector
• Modified all solvers to use matvec/rmatvec instead of @/.H @ to improve performance

v2.1.0

• Added pylops.signalprocessing.DCT, pylops.signalprocessing.NonStationaryConvolve1D,
  pylops.signalprocessing.NonStationaryConvolve2D, pylops.signalprocessing.NonStationaryFilters1D, and
  pylops.signalprocessing.NonStationaryFilters2D operators
• Added pylops.waveeqprocessing.BlendingContinuous, pylops.waveeqprocessing.BlendingGroup, and
  pylops.waveeqprocessing.BlendingHalf operators
• Added kind='datamodel' to pylops.optimization.cls_sparsity.IRLS
• Improved inner working of pylops.waveeqprocessing.Kirchhoff operator significantly
  reducing the memory usage related to storing traveltime, angle, and amplitude tables.
• Improved handling of haxes in pylops.signalprocessing.Radon2D and pylops.signalprocessing.Radon3D operators
• Added possibility to feed ND-arrays to pylops.TorchOperator
• Removed pylops.LinearOperator inheritance and added __call__ method to pylops.TorchOperator
• Removed scipy.sparse.linalg.LinearOperator and added abc.ABC inheritance to pylops.LinearOperator
• All operators are now classes of pylops.LinearOperator type

v2.0.0

PyLops has undergone significant changes in this release, including new LinearOperators, more features, new examples and bugfixes.
To aid users in navigating the breaking changes, we provide the following document MIGRATION_V1_V2.md.

New Features

• Multiplication of linear operators by N-dimensional arrays is now supported via the new dims/dimsd properties.
  Users do not need to use .ravel and .reshape as often anymore. See the migration guide for more information.
• Typing annotations for several submodules (avo, basicoperators, signalprocessing, utils, optimization, waveeqprocessing)
• New pylops.TorchOperator wraps a Pylops operator into a PyTorch function
• New pylops.signalprocessing.Patch3D applies a linear operator repeatedly to patches of the model vector
• Each of Sliding1D, Sliding2D, Sliding3D, Patch2D and Patch3D have an associated slidingXd_design or patchXd_design functions associated with them to aid the user in designing the windows
• FirstDerivative and SecondDerivative, and therefore other derivative operators which rely on the (e.g., Gradient) support higher order stencils
• pylops.waveeqprocessing.Kirchhoff substitutes pylops.waveeqprocessing.Demigration and incorporates a variety of new functionalities
• New pylops.waveeqprocessing.AcousticWave2D wraps the Devito acoutic wave propagator providing a wave-equation based Born modeling operator with a reverse-time migration adjoint
• Solvers can now be implemented via the pylops.optimization.basesolver.Solver class. They can now be used through a functional interface with lowercase name (e.g., splitbregman) or via class interface with CamelCase name (e.g., SplitBregman). Moreover, solvers now accept callbacks defined by the Callbacks interface (see e.g., MetricsCallback).
• Metrics such as mean absolute error (mae), mean squared error (mse) and others
• New pylops.utils.signalprocessing.dip_estimate estimates local dips in an image (measured in radians) in a stabler way than pylops.utils.signalprocessing.dip_estimate did for slopes.
• New pylops.utils.tapers.tapernd for N dimensional tapers
• New wavelets Klauder and Ormsby

Documentation

• Installation has been revamped
• Revamped guide on how to implement a new LinearOperator from scratch
• New guide on how to implement a new solver from scratch
• New tutorials:
  • Solvers (Advanced)
  • Deblending
  • Automatic Differentiation
• New gallery examples:
  • Patching
  • Wavelets

v1.18.3

• Fixed pylops.optimization.basic.lsqr, pylops.optimization.sparsity.ISTA, and
  pylops.optimization.sparsity.FISTA to work with cupy arrays. This change was required
  by how recent cupy versions handle scalars, which are not converted directly into float types,
  rather kept as cupy arrays.
• Fixed bug in pylops.waveeqprocessing.Deghosting introduced in
  commit 7e596d4

v1.18.2

• Added NMO example to gallery
• Extended pylops.basicoperators.Laplacian to N-dimensional arrays
• Added forward kind to pylops.basicoperators.SecondDerivative and
  pylops.basicoperators.Laplacian
• Added chirp-sliding kind to pylops.waveeqprocessing.seismicinterpolation.SeismicInterpolation
• Fixed bug due to the new internal structure of LinearOperator submodule introduced in scipy1.8.0

v1.18.0

• Added NMO example to gallery
• Extended pylops.basicoperators.Laplacian to N-dimensional arrays
• Added forward kind to pylops.basicoperators.SecondDerivative and
  pylops.basicoperators.Laplacian
• Added chirp-sliding kind to pylops.waveeqprocessing.seismicinterpolation.SeismicInterpolation
• Fixed bug due to the new internal structure of LinearOperator submodule introduced in scipy1.8.0

v1.17.0

• Added pylops.utils.describe.describe method
• Added fftengine to pylops.waveeqprocessing.Marchenko
• Added ifftshift_before and fftshift_after optional input parameters in
  pylops.signalprocessing.FFT
• Added norm optional input parameter to pylops.signalprocessing.FFT2D and
  pylops.signalprocessing.FFTND
• Added scipy backend to pylops.signalprocessing.FFT and
  pylops.signalprocessing.FFT2D and pylops.signalprocessing.FFTND
• Added eps optional input parameter in
  pylops.utils.signalprocessing.slope_estimate
• Added pre-commit hooks
• Improved pre-commit hooks
• Vectorized pylops.utils.signalprocessing.slope_estimate
• Handlexd nfft<nt case in pylops.signalprocessing.FFT and
  pylops.signalprocessing.FFT2D and pylops.signalprocessing.FFTND
• Introduced automatic casting of dtype in `pylops.MatrixMult
• Improved documentation and definition of optinal parameters
  of pylops.Spread
• Major clean up of documentation and mathematical formulas
• Major refractoring of the inner structure of pylops.signalprocessing.FFT and
  pylops.signalprocessing.FFT2D and pylops.signalprocessing.FFTND
• Reduced warnings in test suite
• Reduced computational time of test_wavedecomposition in the test suite
• Fixed bug in pylops.signalprocessing.Sliding1D,
  pylops.signalprocessing.Sliding2D and
  pylops.signalprocessing.Sliding3D where the dtype of the Restriction
  operator is inffered from Op
• Fixed bug in pylops.signalprocessing.Radon2D and
  pylops.signalprocessing.Radon3D when using centered spatial axes
• Fixed scaling in pylops.signalprocessing.FFT with real=True to pass the
  dot-test

v1.16.0

• Added pylops.utils.estimators module for trace estimation
• Added x0 in pylops.optimization.sparsity.ISTA and
  pylops.optimization.sparsity.FISTA to handle non-zero initial guess
• Modified pylops.optimization.sparsity.ISTA and
  pylops.optimization.sparsity.FISTA to handle multiple right hand sides
• Modified creation of haxis in pylops.signalprocessing.Radon2D and
  pylops.signalprocessing.Radon3D to allow for uncentered spatial axes
• Fixed _rmatvec for explicit in pylops.LinearOperator._ColumnLinearOperator

v1.15.0

• Added pylops.signalprocessing.Shift operator.
• Added option to choose derivative kind in
  pylops.avo.poststack.PoststackInversion and
  pylops.avo.prestack.PrestackInversion
• Improved efficiency of adjoint of
  pylops.signalprocessing.Fredholm1 by applying complex conjugation
  to the vectors.
• Added vsvp to pylops.avo.prestack.PrestackInversion allowing
  to use user defined VS/VP ratio.
• Added kind to pylops.basicoperators.CausalIntegration allowing
  full, half, or trapezoidal integration
• Fixed _hardthreshold_percentile in
  pylops.optimization.sparsity - Issue #249.
• Fixed r2norm in pylops.optimization.solver.cgls

v1.14.0

• Added pylops.optimization.solver.lsqr solver
• Added utility routine pylops.utils.scalability_test for scalability
  tests when using multiprocessing
• Added pylops.avo.avo.ps AVO modelling option and restructured
  pylops.avo.prestack.PrestackLinearModelling to allow passing any
  function handle that can perform AVO modelling apart from those directly
  available
• Added R-linear operators (when setting the property clinear=False of a
  linear operator). pylops.basicoperators.Real,
  pylops.basicoperators.Imag, and pylops.basicoperators.Conj
• Added possibility to run operators pylops.basicoperators.HStack,
  pylops.basicoperators.VStack, pylops.basicoperators.Block
pylops.basicoperators.BlockDiag,
  and pylops.signalprocessing.Sliding3D using multiprocessing
• Added dtype to vector X when using scipy.sparse.linalg.lobpcg in
  eigs method of pylops.LinearOperator
• Use kind=forward fot FirstDerivative in
  pylops.avo.poststack.PoststackInversion inversion when dealing
  with L1 regularized inversion as it makes the inverse problem more stable
  (no ringing in solution)
• Changed cost in pylops.optimization.solver.cg
  and pylops.optimization.solver.cgls to be L2 norms of residuals
• Fixed pylops.utils.dottest.dottest for imaginary vectors and to
  ensure u and v vectors are of same dtype of the operator