v0.3.2

GeometricMachineLearning v0.3.2

Diff since v0.3.1

Merged pull requests:

• Included tag '*' & Removed python. (#159) (@benedict-96)

v0.3.1

GeometricMachineLearning v0.3.1

Diff since v0.3.0

Merged pull requests:

• Introduced new function include_graphics (#147) (@benedict-96)
• Implemented Tensor-Matrix multiplication for SymmetricMatrix (#148) (@benedict-96)
• Linear symplectic transformer (#150) (@benedict-96)
• Rework manifold docs (#151) (@benedict-96)
• Update of documentation (including docstrings) for custom array types (#152) (@benedict-96)
• Rework optimizer docs (#154) (@benedict-96)
• CompatHelper: add new compat entry for UpdateJulia at version 0.4, (keep existing compat) (#155) (@github-actions[bot])
• CompatHelper: add new compat entry for LazyArrays at version 2, (keep existing compat) (#156) (@github-actions[bot])
• CompatHelper: bump compat for GeometricEquations to 0.17, (keep existing compat) (#157) (@github-actions[bot])
• Online phase tutorial (#158) (@benedict-96)

v0.3.0

GeometricMachineLearning v0.3.0

Diff since v0.2.0

Merged pull requests:

• Grassmann manifold documentation (#112) (@benedict-96)
• Fixed the seed. (#113) (@benedict-96)
• Corrections in docs (#114) (@benedict-96)
• Data loader docs (#115) (@benedict-96)
• Neural network losses (#116) (@benedict-96)
• CompatHelper: bump compat for GeometricEquations to 0.15, (keep existing compat) (#117) (@github-actions[bot])
• CompatHelper: bump compat for GeometricBase to 0.10, (keep existing compat) (#119) (@github-actions[bot])
• Include latex version (#120) (@benedict-96)
• Remove unnecessary dependencies (#121) (@benedict-96)
• CompatHelper: bump compat for GeometricBase to 0.10, (keep existing compat) (#122) (@github-actions[bot])
• Simplified Makefile and created tests for latex docs on ubuntu and macos. (#123) (@benedict-96)
• Adjust to new ann interface (#125) (@benedict-96)
• Fix documenter workflow (#126) (@benedict-96)
• Parallel tensor inverse(s) and Cayley transform(s) for VolumePreservingAttention (#127) (@benedict-96)
• Volume preserving feedforward neural network (vpffnn) (#128) (@benedict-96)
• Loss routines (#129) (@benedict-96)
• Adam with decay (#130) (@benedict-96)
• Transformer integrator (#131) (@benedict-96)
• Update data loader (#132) (@benedict-96)
• Simplify Makefiles (#133) (@benedict-96)
• Docs for volume preserving attention (#134) (@benedict-96)
• Added documentation on pullbacks. (#135) (@benedict-96)
• Volume preserving transformer (#136) (@benedict-96)
• Updated ANN version. (#137) (@benedict-96)
• Symplectic Autoencoder and PSD architecture (#138) (@benedict-96)
• Updated ReducedSystem and renamed to HRedSys (including tutorial) (#140) (@benedict-96)
• Specified compat entries for GeometricSolutions and ForwardDiff. (#146) (@benedict-96)

v0.2.0

GeometricMachineLearning v0.2.0

Diff since v0.1.0

Merged pull requests:

• Transformer utils (#50) (@benedict-96)
• Clean up (#51) (@benedict-96)
• More documentation (#52) (@benedict-96)
• Rnn (#53) (@Theozeud)
• AttentionLayer tests (#54) (@benedict-96)
• Removed the folder custom_neural_networks (#57) (@benedict-96)
• Removed some old zombie functions (relating to device) (#58) (@benedict-96)
• Increase test coverage (#59) (@benedict-96)
• Simplify retractions (#60) (@benedict-96)
• Increase test coverage (by removing zombie files) and added struct DataLoader (#61) (@benedict-96)
• Write optimizer documentation (#62) (@benedict-96)
• Unify sympnet layers (#63) (@benedict-96)
• Increase test coverage for layers (#64) (@benedict-96)
• Add diagram to doc (#66) (@benedict-96)
• Mnist script (#71) (@benedict-96)
• CompatHelper: add new compat entry for ChainRulesTestUtils at version 1, (keep existing compat) (#72) (@github-actions[bot])
• CompatHelper: add new compat entry for Documenter at version 0.27, (keep existing compat) (#73) (@github-actions[bot])
• CompatHelper: add new compat entry for SafeTestsets at version 0.1, (keep existing compat) (#74) (@github-actions[bot])
• Added docs for ROM, added docs for Cayley (and optimized routine), simplified calculation in attention layer. (#76) (@benedict-96)
• Symplectic autoencoder script (#83) (@benedict-96)
• Batch struct (#84) (@benedict-96)
• Update dependencies (#86) (@michakraus)
• Refactoring and documentation of SympNets (#88) (@benedict-96)
• Documentation for optimizer (#89) (@benedict-96)
• Implementation of BFGS optimizer (#90) (@benedict-96)
• Bfgs pr (second attempt) (#92) (@benedict-96)
• Bias for linear layer (#93) (@benedict-96)
• Added comment on parameterlength and extended the corresponding function in ANN. (#94) (@benedict-96)
• Internally now only working with NamedTuples, not with Tuples when dealing with q and p. (#95) (@benedict-96)
• Removed unnecessary image files (pngs) (#96) (@benedict-96)
• Put sampling of arrays into extra test file (#97) (@benedict-96)
• Multiplication and addition tests for custom arrays (#98) (@benedict-96)
• Constructor tests for custom arrays (#99) (@benedict-96)
• Simplify transformer tests (#100) (@benedict-96)
• Increase data loader test coverage (#101) (@benedict-96)
• CompatHelper: bump compat for GPUArrays to 10, (keep existing compat) (#102) (@github-actions[bot])
• Update readme (#103) (@benedict-96)
• Add darkmode for sympnets (#104) (@benedict-96)
• Fixed mnist tutorial (#105) (@benedict-96)
• Removed NLsolve and Optimisers dependency. (#106) (@benedict-96)
• Moved write_ones kernel to utils.jl (#107) (@benedict-96)
• ignore -> ignore_derivatives (#108) (@benedict-96)
• Fix version in the documenter to 1 and fixed cpu allocation problem. (#109) (@benedict-96)
• Add dark mode images for mha and spline and fixed problem with html on github (#111) (@benedict-96)

Closed issues:

• @views for assign_qp and for DataLoader. (#80)
• Fix ChainRulesCore deprecations (#85)
• Missing images in docs (#87)

v0.1.0

GeometricMachineLearning v0.1.0

Merged pull requests:

• Lux implementation of HNNs (#1) (@michakraus)
• CompatHelper: add new compat entry for BandedMatrices at version 0.17, (keep existing compat) (#2) (@github-actions[bot])
• CompatHelper: add new compat entry for NLsolve at version 4, (keep existing compat) (#3) (@github-actions[bot])
• CompatHelper: add new compat entry for NNlib at version 0.8, (keep existing compat) (#4) (@github-actions[bot])
• CompatHelper: add new compat entry for HDF5 at version 0.16, (keep existing compat) (#5) (@github-actions[bot])
• Add a SympNet example script for the pendulum (#6) (@michakraus)
• Fix and finish HNN implementation (#7) (@michakraus)
• Implemented Householder reflections with modifications. (#8) (@benedict-96)
• Merge HNN functions (#9) (@michakraus)
• Implement new architectures and add documentation (#10) (@michakraus)
• CompatHelper: bump compat for StatsBase to 0.34, (keep existing compat) (#11) (@github-actions[bot])
• Implementation of an optimisation structure (#12) (@Theozeud)
• Symplectic householder (#14) (@benedict-96)
• CompatHelper: bump compat for NNlib to 0.9, (keep existing compat) (#15) (@github-actions[bot])
• Training integrator (#16) (@Theozeud)
• CompatHelper: add new compat entry for Flux at version 0.13, (keep existing compat) (#17) (@github-actions[bot])
• Gpu adaptation (#18) (@benedict-96)
• CompatHelper: add new compat entry for CUDA at version 4, (keep existing compat) (#19) (@github-actions[bot])
• CompatHelper: add new compat entry for KernelAbstractions at version 0.9, (keep existing compat) (#20) (@github-actions[bot])
• CompatHelper: add new compat entry for GPUArrays at version 8, (keep existing compat) (#21) (@github-actions[bot])
• Transformer gpu adaptation (#22) (@benedict-96)
• CompatHelper: add new compat entry for ChainRulesTestUtils at version 1, (keep existing compat) (#23) (@github-actions[bot])
• CompatHelper: add new compat entry for ChainRulesCore at version 1, (keep existing compat) (#24) (@github-actions[bot])
• CompatHelper: add new compat entry for ChainRules at version 1, (keep existing compat) (#25) (@github-actions[bot])
• Neural net solution (#26) (@Theozeud)
• Refactor optimizers to be independent of layers (#28) (@michakraus)
• CompatHelper: bump compat for Lux to 0.5, (keep existing compat) (#29) (@github-actions[bot])
• Final fixes to application to ensemble solution (#30) (@michakraus)
• CompatHelper: add new compat entry for GeometricIntegrators at version 0.12, (keep existing compat) (#32) (@github-actions[bot])
• Optimizer tests (#34) (@benedict-96)
• Merge applica abstract nn (#35) (@Theozeud)
• Transformer tests (#39) (@benedict-96)
• Gpu fixes (#40) (@benedict-96)
• CompatHelper: add new compat entry for GeometricBase at version 0.8, (keep existing compat) (#41) (@github-actions[bot])
• CompatHelper: add new compat entry for GeometricEquations at version 0.12, (keep existing compat) (#42) (@github-actions[bot])
• add timer + training method (#43) (@Theozeud)
• Manifold optimizer tests (#44) (@benedict-96)
• Added a simple method so that we can run the dense layer for tensors in parallel. (#45) (@benedict-96)
• CompatHelper: add new compat entry for TimerOutputs at version 0.5, (keep existing compat) (#46) (@github-actions[bot])
• New activation function & transformer script prediction window (#48) (@benedict-96)