v0.6.3

SparseConnectivityTracer v0.6.3

Diff since v0.6.2

• Add DataInterpolations.jl package extension (#178)
• Add LogExpFunctions.jl package extension (#184)
• Add NaNMath.jl package extension (#187)
• Support two-argument atan and log (#185)
• Document limitations of operator overloading utils (#180)
• Reexport ADTypes interface (#182)
• Update developer documentation URLs (#186)
• Reorganize code and update code generation utilities (#179, #183)

Merged pull requests:

• DataInterpolations support (#178) (@SouthEndMusic)
• Reorganize code (#179) (@adrhill)
• Document limitations of operator overloading utils (#180) (@adrhill)
• Minor fixes to documentation (#182) (@adrhill)
• Update code generation utilities (#183) (@adrhill)
• Add LogExpFunctions package extension (#184) (@adrhill)
• Add two-arg atan, log (#185) (@adrhill)
• Rename developer documentation folder (#186) (@adrhill)
• Add NaNMath package extension (#187) (@adrhill)

Closed issues:

• Add package extension for NaNMath.jl (#170)
• Add package extension for LogExpFunctions.jl (#171)

v0.6.2

SparseConnectivityTracer v0.6.2

Diff since v0.6.1

• Return only primal value when applying non-differentiable methods to Dual numbers (#169)
• Increase sparsity for Diagonal inputs (#165)
• Add more methods on round, rand (#162)
• This release brings large updates to the documentation:
  • Document limitations (#175)
  • Document global vs. local patterns (#176)
  • Add "How it works" developer documentation (#174)
  • Add developer documentation on custom overloads (#177)
• Refactor type conversions (#173, #168, #166)
• Make comparisons regular operators (#169)
• Fix Hessian on NNlib activation functions celu, elu, selu, hardswish (#162)
• Fix isless (#161)

Merged pull requests:

• Fix isless and update tests (#161) (@adrhill)
• More methods for round and rand, fix NNlib (#162) (@adrhill)
• Increase sparsity for Diagonal inputs (#165) (@adrhill)
• Fix convert on Dual (#166) (@adrhill)
• Add direct type conversion on Dual (#168) (@adrhill)
• Return only primal when applying non-differentiable methods to Dual (#169) (@adrhill)
• Refactor type conversions (#173) (@adrhill)
• Add "How it works" developer documentation (#174) (@adrhill)
• Document limitations (#175) (@adrhill)
• Document global vs. local sparsity patterns (#176) (@adrhill)
• Document how to add custom overloads (#177) (@adrhill)

Closed issues:

• Document control flow limitations of global tracers (#143)
• Add custom trace_input methods for LinearAlgebra matrix types (#147)
• Add support for round (#152)
• Another problem with isless (#153)
• Add guide on how to add custom overloads (#163)
• More problems regarding round (#164)
• Zero-derivative functions on dual numbers should only return the primal (#167)
• Behavior of zeros (#172)

v0.6.1

SparseConnectivityTracer v0.6.1

Diff since v0.6.0

• Improve the performance of Hessian pattern tracing by an order of magnitude:
  • Internally represent Hessian with dictionaries (#155, #158)
  • Increase performance via symmetry of Hessian (#151)
• Support ComponentArrays (#146)
• Support boolean not (!) (#150)
• Support isless (#149)

Merged pull requests:

• Support ComponentArrays (#146) (@gdalle)
• Support isless (#149) (@adrhill)
• Support boolean not ! (#150) (@adrhill)
• Increase performance via symmetry of Hessian (#151) (@adrhill)
• Bring back dictionaries for Hessian patterns (#155) (@adrhill)
• Rename variables in Hessian tracer code (#156) (@adrhill)
• Smaller benchmarks (#157) (@adrhill)
• Use DictHessianPattern as default tracer pattern (#158) (@adrhill)
• Remove randomness from benchmarks (#159) (@adrhill)
• Rename benchmarks to avoid parsing of tuples (#160) (@adrhill)

Closed issues:

• Support more input types (#28)
• Potential increase in performance via symmetry of Hessian (#34)
• Is GPU support possible for Jacobian tracing? (#67)
• How to handle complex numbers? (#102)
• Add support for ComponentVector (#145)
• Missing negation (#148)
• Return Symmetric Hessians by default? (#154)

v0.6.0

SparseConnectivityTracer v0.6.0

Diff since v0.5.1

• Remove ConnectivityTracer (#140)
• Remove legacy interface (#140)
  • instead of jacobian_pattern(f, x), use jacobian_sparsity(f, x, TracerSparsityDetector())
  • instead of hessian_pattern(f, x), use hessian_sparsity(f, x, TracerSparsityDetector())
  • instead of local_jacobian_pattern(f, x), use jacobian_sparsity(f, x, TracerLocalSparsityDetector())
  • instead of local_hessian_pattern(f, x), use hessian_sparsity(f, x, TracerLocalSparsityDetector())
• Remove overloads on similar to reduce amount of invalidations (#132)
• Fix sparse array construction (#142)
• Add array overloads (#131)
• Generalize sparsity pattern representations (#139, #119)
• Reduce allocations of new tracers (#128)
• Reduce compile times (#119)

Merged pull requests:

• Fix typo in docs (#117) (@adrhill)
• Generalize sparsity pattern representations (#119) (@adrhill)
• Clean up NLPModels test suite (#121) (@gdalle)
• Add CITATION (#125) (@adrhill)
• Fix tracer overloads on exponent function (#126) (@adrhill)
• Slim down benchmark suite (#127) (@adrhill)
• Avoid allocating new tracers (#128) (@adrhill)
• Restructure code and rename operator traits (#129) (@adrhill)
• Add array overloads (#131) (@adrhill)
• Remove overloads on similar (#132) (@adrhill)
• Add shared Hessian tracer à la Walther (#135) (@gdalle)
• Put benchmark boilerplate into subpackage (#136) (@gdalle)
• Introduce sparsity patterns (#139) (@adrhill)
• Remove ConnectivityTracer and legacy API (#140) (@adrhill)
• Fix sparse array construction (#142) (@adrhill)

Closed issues:

• Add isempty flag to tracers (#80)
• Shared sets and sets with shared objects (#107)
• StackOverflow when creating SparseMatrixCSC with tracers as values (#108)
• Generalize first- and second-order information in HessianTracer (#114)
• Add AbstractArray{<:AbstractTracer} methods for common LinearAlgebra functions (#115)
• [documentation] typo in a Tracer (#116)
• Benchmarking against JuMP: faster on Jacobians, slower on Hessians (#118)
• ADNLPModels test suite fails (#120)
• SnoopCompile investigation (#122)
• Remove overloads on similar (#123)
• Missing sparse constructor overloads (#141)

v0.5.1

SparseConnectivityTracer v0.5.1

Diff since v0.5.0

Merged pull requests:

• Add NLPModels testing (#83) (@gdalle)
• Allow more generic hessian set types (#103) (@gdalle)
• Get rid of UInt64 (#109) (@gdalle)
• Fix float, fix constructor tests (#110) (@adrhill)
• Measure code coverage of test directory (#111) (@adrhill)
• Fix failing benchmarks (#112) (@gdalle)
• Fix real, improve code coverage (#113) (@adrhill)

Closed issues:

• Add optimization problems to "real-world" tests (#69)
• Allow more Hessian set types than just Set{Tuple{Int,Int}} (#105)

v0.5.0

SparseConnectivityTracer v0.5.0

Diff since v0.4.2

Merged pull requests:

• Support ifelse (#94) (@adrhill)
• Restructure benchmarks, add Hessian test cases (#104) (@gdalle)
• Fix 3-argument similar and add tests on type conversions (#106) (@adrhill)

Closed issues:

• Add benchmarks for Dual and Hessian tracers (#88)
• Global sparsity detection with control flow (#90)

v0.4.2

SparseConnectivityTracer v0.4.2

Diff since v0.4.1

Merged pull requests:

• Make AbstractTracer subtype of Real (#92) (@adrhill)
• Add back global Jacobian tests (#93) (@adrhill)
• Limit Dual to Real primals, replace all Number with Real (#95) (@gdalle)
• Add stdlib compat <0.0.1 (#97) (@gdalle)
• Support Hessians on non-tracer outputs (#99) (@gdalle)
• Constrain type of primal value in Dual (#101) (@adrhill)

Closed issues:

• Make AbstractTracer a subtype of Real, not Number (#91)
• Compat bounds for stdlibs (#96)
• Support functions that don't depend on the input (#98)
• Weird ambiguity on 1.6 but not on 1.10 (#100)

v0.4.1

SparseConnectivityTracer v0.4.1

Diff since v0.4.0

Merged pull requests:

• Replace operator symbols with operator objects (#71) (@gdalle)
• Add conj, real, transpose, adjoint on numbers (#75) (@gdalle)
• Revamp classification tests (#76) (@gdalle)
• Put overloading mechanism into functions (#77) (@gdalle)
• Add influence classification for ConnectivityTracer (#78) (@gdalle)
• Test on 1.6 too (#79) (@gdalle)
• Add TracerLocalSparsityDetector (#81) (@adrhill)
• Add SpecialFunctions extension (#82) (@gdalle)
• Fix dual HessianTracer on exponent (#84) (@adrhill)
• Add NNlib extension (#85) (@adrhill)
• Split Connectivity- and GradientTracer tests (#86) (@adrhill)
• Remove unused interpolation (#87) (@adrhill)
• Add back interpolation (#89) (@adrhill)

Closed issues:

• Support "Hessians" for vector-valued functions (#29)
• Add tests for local derivatives (#66)
• Support for linear algebra (#68)
• Local sparsity support with ADTypes (#72)
• Support dot products (#73)

v0.4.0

SparseConnectivityTracer v0.4.0

Diff since v0.3.3

New features

This release adds dual-number tracers that compute local sparsity patterns and are compatible with functions that require primal value information (e.g. iszero, >, ==).

These can be used via three new functions:

• local_jacobian_pattern
• local_hessian_pattern
• local_connectivity_pattern

Merged pull requests:

• Refactor code to match theory (#59) (@adrhill)
• Add local tracers (#65) (@adrhill)
• Support Flux Conv layers with ReLU activation functions (#70) (@adrhill)

Closed issues:

• Return both Hessian and Jacobian patterns (#30)
• Add local tracers (#56)
• Refactor following theory (#57)
• Operators with local sparsity > global sparsity (#64)
• Support for linear algebra (#68)

v0.3.3

SparseConnectivityTracer v0.3.3

Diff since v0.3.2

Merged pull requests:

• Bump julia-actions/cache from 1 to 2 (#49) (@dependabot[bot])
• Bump actions/cache from 3 to 4 (#60) (@dependabot[bot])
• Sparse matrices with Int indices (#62) (@gdalle)

Closed issues:

• Switch to Int by default instead of UInt (#61)