Bristlecone - v3.0.0

• FEATURE: Whole library and language is Tensor-based using DiffSharp.
• FEATURE: Units of measure in language (model definitions).
• FEATURE: Inverse functions in language using Newton–Raphson or bisection.
• FEATURE: Fast grid-based inverse for root finding
• FEATURE: Heteroscedacity in simple likelihood functions
• BUGFIX: Amoeba solver correctly takes account of dimensionality

What's Changed

• Ability to change date modes in time-series types by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/30
• Tensor-based DSL with units of measure for model equations by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/32
• Package docs and dependencies for release by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/33
• Update readme by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/34
• JOSS manuscript and support for model temporal resolutions by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/37
• Fix documentation deployment by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/38
• Performance improvements for Inverse nodes by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/39
• Unit-typed, stronger likelihood functions and test scaffolding by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/41
• Metadata + bivariate gauss fix by @AndrewIOM in https://github.com/AndrewIOM/bristlecone/pull/43

Full Changelog: https://github.com/AndrewIOM/bristlecone/compare/v2.0.0...v3.0.0

Biosphere - Forest Modeling and Analysis - F#
Published by AndrewIOM 4 months ago

Bristlecone - v2.0.0

Includes new language for writing models, functions for constructing multiple-hypothesis testing, and tests and benchmarks.

• FEATURE: Domain Specific Language (DSL) for scaffolding models.
• FEATURE: User-definable RNG seeds.
• FEATURE: Scaffold many models into competing hypotheses.
• FEATURE: One-step-ahead analysis of model performance.
• FEATURE: Benchmark suite of common test functions for optimisation routines
• FEATURE: Performance tuning of optimisers. For example, Filzbach 94% faster.
• FEATURE: Documentation website uses latest fsdocs and includes fully worked examples.
• BREAKING: We recommend using the DSL to construct models. Underlying F# record type signatures have changed and will break version 1.x scripts.
• BUGFIX: Amoeba functions respect end conditions.

Biosphere - Forest Modeling and Analysis - F#
Published by AndrewIOM about 2 years ago