v0.9

calibrar 0.9

• new optim2() equivalent to stats::optim() but with parallel computation of numerical gradients.

• new optimh() wrapping heuristic methods with the same syntax of stats::optim().

• the calibrate() function implements the restart functionality for the Rvmmin method too, useful for the optimization of deterministic functions with long runtime.

• Improved methods for visualization of results.

• All optimization methods available in calibrate() can use functions reading and writing from the disk.

• Functioncalibrate() can use a different method for each estimation phase.

• calibrate() is a generic now.

• Automatic stopping criteria for the AHR-ES method:

      - 0: maxit/maxgen only
      - 1: 1 OR max step reduction
      - 2: relative tolerance on value (smoothing for AHR-ES)
      - 3: maximum number of generations without improvement of `reltol`.
  

• Automatic testing using testthat package.

• Automatic support to optimize functions produced with the TMB package, via a method for calibrate().

• getCalibrationInfo(), createObjectiveFuction() and getObservedData() are defunct now.

v0.3

calibrar v0.3.0

• new optimization methods available in calibrate(): 'LBFGSB3', 'hjn', 'CMA-ES', 'genSA', 'DE', 'soma', 'genoud', 'PSO', 'hybridPSO', 'mads'.
• fine control of numerical gradient computations, including parallelization.
• replicates argument for stochastic functions
• several minor bugs fixed
• getCalibrationInfo(), createObjectiveFuction() and getObserved() data are replaced and deprecated and replaced by calibration_setup(), calibration_objFn() and calibration_data().
• spline_par() function to simplify the estimation of smooth time-varying parameters.
• several minor bugs fixed

This version can be installed from the osmose-model drat repository:

install.packages("calibrar", repo="https://osmose-model.github.io/drat/")

v0.2.0

calibrar v0.2.0

First version available at CRAN: https://cran.r-project.org/web/packages/calibrar/index.html

• Handles lists as parameter argument.
• 18 optimization methods available now.

Automated parameter estimation for complex (ecological) models in R. This package allows the parameter estimation or calibration of complex models, including stochastic ones. It is a generic tool that can be used for fitting any type of models, especially those with non-differentiable objective functions. It supports multiple phases and constrained optimization. It implements maximum likelihood estimation methods and automated construction of the objective function from simulated model outputs.

First release

Calibration of Ecological using Evolutionary Algorithms

The calibration of complex ecological models is a challenging optimization task, with a notable lack of tools for the calibration of stochastic models. The _calibrar_ package is a new R package for the calibration of stochastic ecological models, including Individual Based Models (IBMs). It is a generic tool that can be used for any type of model, especially those with non-differentiable objective functions. _calibrar_ supports multiple phase calibrations and constrained optimization. It implements maximum likelihood estimation methods and automated construction of the objective function from simulated model outputs.