ionBench 0.5.0

ionBench v0.5.0-alpha

Reworked the tracker class (which breaks compatibility). New rProp optimisers from pints. Other bug fixes and improvements.

ionBench 0.4.3

ionBench v0.4.3-alpha

Further bug fixes and minor improvements in the optimisers, tracker class and benchmarker class

ionBench 0.4.2

ionBench v0.4.2-alpha

Some more minor fixes for results gathering. Fixed a bug in pints optimisers for Moreno. Corrected the stochastic search optimiser.

ionBench 0.4.1

ionBench v0.4.1-alpha

Hot fix for a bug in steady states (mainly relevant for Moreno but also affects staircase MM)

ionBench 0.4.0

ionBench v0.4.0-alpha

There are a massive number of changes made in this version. Large refactoring changes. Many bugs fixes in both optimisers and problems. New definitions for the staircase problems. New features like rate bounds can be applied via modification to any problem-optimisation combination, penalty function for bound violations, consistent assumptions in pso optimisers.

ionBench 0.3.4

ionBench v0.3.4-alpha

Fixed initial state bug that was still in v0.3.3

ionBench 0.3.3

ionBench v0.3.3-alpha

Bug fix in the steady state setting. Previously caused crashes for markov model.

ionBench 0.3.2

ionBench v0.3.2-alpha

Updates to the profile likelihood code. Bug fixes to ensure it runs with the new lm scipy method (requires a gradient calculator in the ProfileManager) and removed noise from the staircase Markov model problem to improve parameter identifiability

ionBench 0.3.1

ionBench v0.3.1-alpha

Tracker now saves the times for model solves, separated by cost and gradient solves. Improved the scripts to run all optimisers and review their outputs.

ionBench 0.3.0

ionBench v0.3.0-alpha

Added many more optimisers. Introduced lists of approaches to loop through and run everything. Added more features to the Tracker (now tracks time for model solves, reports if model was solve multiple times at the same parameters, and tracks best parameters). Added more tests for the optimisers through a new problem specifically for ensuring all optimisers can optimise a simple problem. All problems now simulate from steady state for current parameters, rather than steady state for true parameters. Gradient calculator doesn't yet account for this.