v0.2.4

NeuralLyapunov v0.2.4

Diff since v0.2.3

Changes

• Removes assumption that all functions will <: Function; improves support for ODEInputFunction in the process
• Some changes to CI, including formatting and downgrade checks
• Corrects some formatting issues
• Adds some tests

Merged pull requests:

• Update CI workflows for downgrade v2 (issue #1076) (#71) (@ChrisRackauckas)
• Update code formatting with JuliaFormatter (#72) (@ChrisRackauckas)
• Adds GPU test of benchmark with policy search and ODESystem (#73) (@nicholaskl97)
• Revert CI downgrade update back to v1 from v2 (#75) (@nicholaskl97)
• Remove references to Function (#77) (@nicholaskl97)

Closed issues:

• Register v0.2.3 (#70)

v0.2.3

NeuralLyapunov v0.2.3

Diff since v0.2.2

Merged pull requests:

• Update NeuralLyapunov tests to use NeuralLyapunovProblemLibrary (#68) (@nicholaskl97)
• Fix benchmark mass matrix problem (#69) (@nicholaskl97)

Closed issues:

• Register v0.2.2 (#67)

v0.2.2

NeuralLyapunov v0.2.2

Diff since v0.2.1

Merged pull requests:

• Bump crate-ci/typos from 1.33.1 to 1.34.0 (#65) (@dependabot[bot])
• Adds init_states option for benchmark, along with test for GPU benchmarking (#66) (@nicholaskl97)

Closed issues:

• Register v0.2.1 (#64)

v0.2.1

New feature: adds AdditiveLyapunovNet and MultiplicativeLyapunovNet.

Additionally bumps compat to include ForwardDiff 1 and updates some GitHub actions, mostly around CompatHelper.

What's Changed

• CompatHelper: bump compat for ForwardDiff to 1, (keep existing compat) by @github-actions in #54
• CompatHelper expansion by @nicholaskl97 in #58
• Fixes failed CompatHelper expansion by @nicholaskl97 in #59
• Bump crate-ci/typos from 1.32.0 to 1.33.1 by @dependabot in #62
• Adds AdditiveLyapunovNet and MultiplicativeLyapunovNet by @nicholaskl97 in #63

Full Changelog: v0.2.0...v0.2.1

v0.2.0

NeuralLyapunov v0.2.0

What's Changed

• Documentation website added! Including demos in #28, #27, #16
• Added training option for region of attraction approximation in #18, #29, #42
• Added support for policy search in #21
• Added benchmarking function in #31, #43, #44, #45, #48, #46
• Added a problem library package (NeuralLyapunovProblemLibrary) in the same repository in #41
• Various interface changes and minor performance improvements by in #24, #28, #30, #53
• More stable tests via StableRNGs by in #37
• GitHub Actions CI testing has been refined and expanded to include formatting and compatibility checking (and compatibility requirements have been updated accordingly) in #47, #49, #52, #17, #51, #33, #38, #39, #50, #32, #15, #14, #23, #25, #26
• Added BuildKite testing by in #35, #36
• Added test of GPU support in #34

Full Changelog: v0.1.0...v0.2.0

NeuralLyapunov.jl v0.1.0

This version supports changes made to the SciMLBase API and now features testing with Pkg.
Demos have become tests.
Additionally, documentation generation has been added

Full Changelog: v0.0.2...v0.1.0

NeuralLyapunov.jl v0.0.2

NeuralLyapunov.jl is a library for searching for neural Lyapunov functions by setting up a NeuralPDE.jl PDESystem.

In v0.0.2, dynamics now are expected to be defined as $\frac{dx}{dt} = f(x,p,t)$ rather than $\frac{dx}{dt} = f(x)$. This adds greater consistency with DifferentialEquations.jl and allows for parameterization of the dynamics. Since there is still no support for time-varying Lyapunov functions, the ODE must be autonomous and only t = 0.0 will be used.

NeuralLyapunov.jl v0.0.1

This is the initial public version of NeuralLyapunov.jl, a library for searching for neural Lyapunov functions by setting up a NeuralPDE.jl PDESystem.