v0.3.3

RobustNeuralNetworks v0.3.3

Diff since v0.3.2

Merged pull requests:

• 147 small bug in docs (#148) (@nic-barbara)

Closed issues:

• Bump to v0.3.2 (#142)
• Constraint the sensitivity (gain) around a point (#145)
• Specify different limit per pair in a multi-input multi-output neural network (#146)
• Small bug in docs (#147)

v0.3.2

RobustNeuralNetworks v0.3.2

Diff since v0.3.1

Merged pull requests:

• Minor updates to docs (#137) (@nic-barbara)
• Updated broken links (#138) (@nic-barbara)
• Issue 136 139 (#140) (@MrstupidJ)
• Bump to v0.3.2 (#141) (@nic-barbara)

Closed issues:

• Bump to v0.3.1 (#135)
• The product of passivity indices does have a theoretical upper bound. (#136)
• Array mutating problem when calculating inverse of (I + M) (#139)

v0.3.1

RobustNeuralNetworks v0.3.1

Diff since v0.3.0

Merged pull requests:

• Changed naming of cayley norm function to improve readability (#127) (@nic-barbara)
• Tidying up a few changes (#128) (@nic-barbara)
• Fixed bug in parameter extraction of SandwichFC (#131) (@nic-barbara)
• Hybridpassive REN step 1: Incrementally Strictly Output Passive (ISOP) model (#132) (@MrstupidJ)
• Fixed a bug in the docs and removed LiveServer.jl as a dependency (#134) (@nic-barbara)

Closed issues:

• Use Flux.Recur to run nonlinear simulation (#67)
• Bump to v0.3.0 (#125)
• Replace norm_cayley with cayley_norm (#126)
• Bug in SandwichFC (#130)

v0.3.0

RobustNeuralNetworks v0.3.0

Diff since v0.2.3

Closed issues:

• Improve package loading speed (#75)
• Switch autograd to ChainRules.jl instead of using Zygote.jl adjoints (#94)
• Improve speed with back-propagation (#100)
• Basic GPU Support (#105)
• Bump to v0.2.3 (#111)
• Docs for observer example are outdated (#112)
• Add GPU support to documentation (#122)

Merged pull requests:

• Updated random seeds and observer documentation (#113) (@nic-barbara)
• Change REN adjoints to ChainRules.jl rrule (#114) (@nic-barbara)
• CompatHelper: add new compat entry for ChainRulesCore at version 1, (keep existing compat) (#115) (@github-actions[bot])
• Attempt to improve package loading speeds (#116) (@nic-barbara)
• Improve speed with back propagation (#117) (@nic-barbara)
• Basic GPU Support (#118) (@nic-barbara)
• Feature/gpu support debug (#121) (@nic-barbara)
• Added docs for basic GPU support (#123) (@nic-barbara)
• Bumped to version 0.3 (#124) (@nic-barbara)

v0.2.3

RobustNeuralNetworks v0.2.3

Diff since v0.2.2

Closed issues:

• Change name of SandwichFCLayer to DenseLBDNLayer (#74)
• Use weight initialisations from Flux.jl (#76)
• Bump to version v0.2.2 (#96)
• Error taking gradients through REN with nx = 0 and/or nv = 0 (#99)
• Restrictive initialisation of RENs (#102)
• Dynamics in RL and Observer Examples is Incorrect (#106)

Merged pull requests:

• Example/animated contraction (#97) (@nic-barbara)
• Bugfix for Linear RENs and Type Stability (#98) (@nic-barbara)
• Bugfix: error when taking gradients with nv = 0 or nx = 0 (#101) (@nic-barbara)
• Feature/rand init (#103) (@nic-barbara)
• CompatHelper: bump compat for Flux to 0.14, (keep existing compat) (#104) (@github-actions[bot])
• Bugfix: box dynamics (#107) (@nic-barbara)
• Updated links and references to papers (#108) (@nic-barbara)
• Docs/update (#109) (@nic-barbara)
• Version bump and Documentation Updates (#110) (@nic-barbara)

v0.2.2

RobustNeuralNetworks v0.2.2

Diff since v0.2.1

Closed issues:

• Bump to version v0.2.1 (#87)
• Make the Lipschitz bound an optional learnable parameter (#89)

Merged pull requests:

• Learnable Lipschitz parameter (#90) (@nic-barbara)
• Switched to log gamma for storing Lipschitz bound in LBDN, and added … (#91) (@nic-barbara)
• Modifying content while writing JuliaCon paper (#92) (@nic-barbara)
• Added citations and cleaned up repository following arXiv submission. (#93) (@nic-barbara)
• Bumped to version 0.2.2 (#95) (@nic-barbara)

v0.2.1

RobustNeuralNetworks v0.2.1

Diff since v0.2.0

Closed issues:

• Missing 1/alpha_bar^2 here? (#17)
• Bump to version v0.2.0 (#73)
• Licence file reference missing in source files (#82)

Merged pull requests:

• Example/pde obsv (#78) (@yuruizhang06)
• Added code owners file (#79) (@jclinton830)
• Example/lbdn rl (#80) (@nic-barbara)
• Licence file reference for all source files (#83) (@jclinton830)
• License file reference in all source files (#84) (@jclinton830)
• Tidy code (#85) (@nic-barbara)
• Example/block observer (#86) (@nic-barbara)
• Bumped to v0.2.1 (#88) (@nic-barbara)

v0.2.0

RobustNeuralNetworks v0.2.0

Diff since v0.1.0

Closed issues:

• Package Registration (#52)
• Array mutation in solve_tril_layer() (#55)
• Polar parameterisation for REN is incorrect (#57)
• Trainable parameters should be specified with @functor (#61)
• Bug in subtypes of AbstractRENParams when not using polar_param (#63)
• Images not appearing in documentation (#69)

Merged pull requests:

• Optimization of the solve_tril_layer function (#54) (@nic-barbara)
• Speed optimization (#56) (@EliottEccidio)
• Fixed polar parameterisation bug (#58) (@nic-barbara)
• Feature/lbdn (#59) (@nic-barbara)
• Switched LBDN to @functor macro, changed examples (#60) (@nic-barbara)
• Refactor/flux interface (#62) (@nic-barbara)
• Bugfix/polar param (#64) (@nic-barbara)
• Feature/lbdn examples (#65) (@nic-barbara)
• Feature/sandwich layer (#66) (@nic-barbara)
• Example/youla ren (#68) (@nic-barbara)
• Testing issue with relative paths in HTML (#70) (@nic-barbara)
• 69 images not appearing in documentation (#71) (@nic-barbara)
• Fixed bug and updated project version (#72) (@nic-barbara)

v0.1.0

RobustNeuralNetworks v0.1.0

Closed issues:

• Default Q, S, R in GeneralRENParams (#16)
• General REN not passing test for behavioural constraint with Q, S, R (#18)
• passive ren (#23)
• Merge output layer into DirectParams (#24)

Merged pull requests:

• Feature/basic functionality (#1) (@nic-barbara)
• Feature/general ren params (#2) (@nic-barbara)
• CompatHelper: add new compat entry for DocStringExtensions at version 0.9, (keep existing compat) (#3) (@github-actions[bot])
• CompatHelper: add new compat entry for MatrixEquations at version 2, (keep existing compat) (#4) (@github-actions[bot])
• CompatHelper: add new compat entry for BenchmarkTools at version 1, (keep existing compat) (#5) (@github-actions[bot])
• CompatHelper: add new compat entry for CUDA at version 3, (keep existing compat) (#6) (@github-actions[bot])
• CompatHelper: add new compat entry for Zygote at version 0.6, (keep existing compat) (#7) (@github-actions[bot])
• CompatHelper: add new compat entry for Flux at version 0.13, (keep existing compat) (#8) (@github-actions[bot])
• Feature/nondiff wrapper (#9) (@nic-barbara)
• Feature/tidy code (#10) (@nic-barbara)
• Feature/lipschitz ren params (#19) (@nic-barbara)
• Issue/general ren behavioural constraints (#20) (@nic-barbara)
• Feature/dev wrappers (#21) (@nic-barbara)
• Changed unit tests to check dissipation inequality with V_{t+1} - V_t on LHS (#22) (@nic-barbara)
• Feature/merge output direct (#25) (@nic-barbara)
• Feature/lbdn (#27) (@ppauli)
• Feature/passive ren (#29) (@MrstupidJ)
• Changed name of package to RobustNeuralNetworks.jl (#33) (@nic-barbara)
• CompatHelper: add new compat entry for MatrixEquations at version 2, (keep existing compat) (#35) (@github-actions[bot])
• CompatHelper: bump compat for CUDA to 4, (keep existing compat) (#37) (@github-actions[bot])
• CompatHelper: add new compat entry for Flux at version 0.13, (keep existing compat) (#38) (@github-actions[bot])
• CompatHelper: add new compat entry for Zygote at version 0.6, (keep existing compat) (#40) (@github-actions[bot])
• Feature/documentation (#42) (@nic-barbara)
• Adding tokens to docs (#43) (@nic-barbara)
• Feature/documentation (#45) (@nic-barbara)
• Docs should now update on merge to master (#46) (@nic-barbara)
• Feature/documentation (#47) (@nic-barbara)
• Feature/documentation (#48) (@nic-barbara)
• Removed unused packages from Project.toml (#49) (@nic-barbara)
• Feature/documentation (#50) (@nic-barbara)
• Added links in examples (#51) (@nic-barbara)