Skip to content

Latest commit

 

History

History
59 lines (52 loc) · 2.93 KB

CHANGELOG.md

File metadata and controls

59 lines (52 loc) · 2.93 KB

v0.6.0:

  • Many minor bugfixes and many corrections to the documentation examples.
  • Dataset now always converts the input into vector of SVectors for better interaction with NearestNeighbors.jl
  • [BREAKING] : timeseries function was renamed to trajectory.
    • Now the documentation is clear: trajectory means a set of one-dimensional timeseries (or a multi-dimensional timeseries), while timeseries always means a one-dimensional timeseries
  • Removed function perform_non0hist, as it had no reason to exist.

v0.5.0:

  • Bugfix where periodicorbits was not considering disttol keyword.
  • Removed rescale argument from lyapunov. This is technically breaking.
  • Bugfixes in lyapunov.
  • Improved default parameters in lyapunov as well as type-stability.

v0.4.5:

  • Major bugfix in the lyapunov function where the time averaging was actually not happening correctly for discrete systems, resulting in slight inaccuracy.
  • Added option to choose whether you want the convergence timeseries for lyapunov, for both discrete and continuous systems.
  • Added option to supply your own functions to rescale and initialize the test state in lyapunov for both continuous and discrete systems.
  • Bugfix related to estimate_delay.

v0.4.0:

  • Increased speed of Standard Map system (by using while instead of mod).
  • Added a huuuge method due to Schmelcher & Diakonos which finds stable and unstable fixed points of any order for discrete maps!
  • Added append! method for Dataset.

v0.3.0:

  • Added method to estimate optimal delay time estimate_delay.
  • Added abstract AbstractDataset supertype.
  • Completely changed the way Reconstruction is done.
  • Added more methods, interaction and smart indexing related to the (new) Reconstruction object. See the new docstring of reconstruct for usage.
  • Added a new method that enables the calculation of the maximum Lyapunov exponent from a numeric timeseries (like an experimental recording). This method (called numericallyapunov) operates on a reconstruction, using 4 different algorithms. See the docstring of numericallyapunov for more info!
  • Added double_pendulum to famous systems.

v0.2.0: Major Improvements and new features

  • Improved algorithm for deducing best boxsizes (using min_pairwise_distance)
  • All numerical data is now structured and expected to be in the Dataset Type, which by itself defines a new interface (identical to matrix).
  • Major speed improvements in lyapunov function calculation.
  • Major speed improvements in histogram computation.
  • A new qr method for small matrices, faster than Base.qr (see qr_sq).
  • Reconstruction interface.
  • New and better docs.
  • Better integration of SVectors in the package all around.
  • Better default keywords for attractor dimension and lyapunov exponents.
  • Equations of motion on all famous systems.

v0.1.0: First release

Normal changelog will be kept from this point onwards.