Generates a list of parameters to be used in a pyDarwin run.
create_pyDarwinOptions(
author = "",
project_name = NULL,
algorithm = c("GA", "EX", "GP", "RF", "GBRT", "PSO"),
GA = pyDarwinOptionsGA(),
PSO = pyDarwinOptionsPSO(),
random_seed = 11,
num_parallel = 4,
num_generations = 6,
population_size = 4,
num_opt_chains = 4,
exhaustive_batch_size = 100,
crash_value = 99999999,
penalty = pyDarwinOptionsPenalty(),
downhill_period = 2,
num_niches = 2,
niche_radius = 2,
local_2_bit_search = TRUE,
final_downhill_search = TRUE,
search_omega_blocks = FALSE,
search_omega_bands = FALSE,
individual_omega_search = TRUE,
search_omega_sub_matrix = FALSE,
max_omega_sub_matrix = 4,
model_run_timeout = 1200,
model_run_priority_class = c("below_normal", "normal"),
postprocess = pyDarwinOptionsPostprocess(),
keep_key_models = TRUE,
use_saved_models = FALSE,
saved_models_file = "{working_dir}/models0.json",
saved_models_readonly = FALSE,
remove_run_dir = FALSE,
remove_temp_dir = TRUE,
use_system_options = TRUE,
model_cache = "darwin.MemoryModelCache",
model_run_man = c("darwin.LocalRunManager", "darwin.GridRunManager"),
engine_adapter = c("nlme", "nonmem"),
working_dir = NULL,
data_dir = NULL,
output_dir = "{working_dir}/output",
temp_dir = NULL,
nlme_dir = "C:/Program Files/Certara/NLME_Engine",
gcc_dir = "C:/Program Files/Certara/mingw64",
nmfe_path = NULL,
rscript_path = file.path(R.home("bin"), "Rscript"),
nlme_license = NULL,
generic_grid_adapter = pyDarwinOptionsGridAdapter(),
...
)A list of pyDarwin options.
Character string: The name of the author.
Character string (optional): The name of the project. If not specified, pyDarwin will set its value to the name of the parent folder of the options file.
Character string: One of EX, GA, GP, RF, GBRT, PSO. See section Details below for more information.
List: Options specific to the Genetic Algorithm (GA). See
pyDarwinOptionsGA().
List: Options specific to the Particle Swarm Optimization (PSO).
See pyDarwinOptionsPSO().
Positive integer: Seed for random number generation.
Positive integer: Number of models to execute in parallel, i.e., how many threads to create to handle model runs.
Positive integer: Number of iterations or generations of the search algorithm to run. Not used/required for EX.
Positive integer: Number of models to create in every generation. Not used/required for EX.
Positive integer: Number of parallel processes to perform the "ask" step (to increase performance). Required only for GP, RF, and GBRT.
Positive integer: Batch size for the EX (Exhaustive Search) algorithm.
Positive real: Value of fitness or reward assigned when model output is not generated. Should be set larger than any anticipated completed model fitness.
List: Options specific to the penalty calculation. See
pyDarwinOptionsPenalty().
Integer: How often to run the downhill step. If < 1, no periodic downhill search will be performed.
Integer: Used for GA and downhill. A penalty is assigned
for each model based on the number of similar models within a niche radius.
This penalty is applied only to the selection process (not to the fitness
of the model). The purpose is to ensure maintaining a degree of diversity
in the population (integer). num_niches is also used to select the number
of models that are entered into the downhill step for all algorithms,
except EX.
Positive real: The radius of the niches. The niche radius is used to define how similar pairs of models are. This is used to select models for the Local search, as requested, and to calculate the sharing penalty for Genetic Algorithm.
Logical: Whether to perform the two-bit local search. The two-bit local search substantially increases the robustness of the search. All downhill local searches are done starting from num_niches models.
Logical: Whether to perform a local search (1-bit and 2-bit) at the end of the global search.
Logical: whether to perform search for block
omegas. Used only when engine_adapter == 'nlme'.
Logical: whether to perform search for band
omegas. Used only when engine_adapter == 'nonmem'.
Logical: If set, every search block will be
handled individually: each block will have a separate gene and max omega
search length (either calculated or set explicitly in the options). If set
to FALSE, all search blocks will have the same pattern of block omegas.
Default is TRUE.
Logical: set to true to search omega
submatrix. Default is FALSE.
Integer: Maximum size of sub matrix to use in search. Default is 4.
Positive real: Time (seconds) after which the execution will be terminated, and the crash value assigned.
Character string (Windows only): Priority
class for child processes that build and run models, as well as run the R
postprocess script. Options are below_normal and normal. below_normal
is recommended to maintain user interface responsiveness.
List: Options specific to postprocessing. See
pyDarwinOptionsPostprocess()
Logical: Key model is the best model in population
(generation). Such models may be a subject of interest when the search is
analyzed, so they should be saved separately with all their output. Default
is TRUE
Logical: Whether to restore saved Model Cache from
file. Default is FALSE.
Character string: The file from which to restore Model Cache. Will only have an effect if use_saved_models is set to true. By default, the cache is saved in {working_dir}/models.json and cleared every time the search is started. To use saved runs, rename models.json or copy it to a different location.
Logical: Do not overwrite the saved_models_file
content. Default is FALSE.
Logical: If TRUE, will delete the entire model run
directory, otherwise - only unnecessary files inside it. Default is
FALSE.
Logical: Whether to delete the entire temp_dir after
the search is finished or stopped. Doesn't have any effect when the search
is run on a grid. Default is TRUE.
Logical: Whether to override options with
environment-specific values. Default is TRUE.
Character string: ModelCache subclass to be used.
Currently, there are only darwin.MemoryModelCache and
darwin.AsyncMemoryModelCache. You can create your own and use it (e.g., a
cache that stores model runs in a database). The name is quite arbitrary
and doesn't have any convention/constraints.
Character string: ModelRunManager subclass to be used.
Currently, there are only darwin.LocalRunManager and
darwin.GridRunManager.
Character string: ModelEngineAdapter subclass to be
used. Currently only nlme (default) and nonmem are available.
Character string: The project's working directory, where
all the necessary files and folders are created. By default, it is set to
<pyDarwin home>/{project_stem}, where {project_stem} is a file system
friendly representation of the project name in a way that it will be easy
to manage as a folder name where all non-letters and non-digits are
replaced with underscores.
Character string: Directory where datasets are located. Must
be available for individual model runs. Default in pyDarwin if not given:
{project_dir}.
Character string: Directory where pyDarwin output will be
placed. Default is {working_dir}/output.
Character string: Parent directory for all model runs' run
directories, i.e., where all folders for every iteration are located.
Default in pyDarwin if not given: {working_dir}/temp.
Character string: Directory where the NLME Engine is
installed/unzipped. Default: C:/Program Files/Certara/NLME_Engine.
Used only when engine_adapter == 'nlme'.
Character string: Directory where the Mingw-w64 compiler (gcc)
is installed. Default: C:/Program Files/Certara/mingw64 for Windows and
gcc version found by which gcc on Linux.
Used only when engine_adapter == 'nlme'.
Character string: Directory where NONMEM is installed.
Used only when engine_adapter == 'nonmem'.
Character string: Path to the Rscript executable. By default, it is obtained with R.home("bin").
Character string (optional): Path to the license file. If not provided, pyDarwin will set its value to PhoenixLicenseFile (only for current Python session).
List: Options specific to the grids. See
pyDarwinOptionsGridAdapter()
Additional parameters: Other arguments not explicitly defined in the function's signature are allowed and will be stored in the options list. See pyDarwin documentation.
The algorithm parameter specifies the type of search algorithm to be used in the pyDarwin optimization process. It determines the strategy and approach used to explore the search space and find the optimal solution. The following are the available options for the algorithm parameter.
"EX" (Exhaustive Search Algorithm): The exhaustive search algorithm is a simple and straightforward method to explore the entire search space systematically. The search space is initially represented as a string of integers, one for each dimension. The algorithm exhaustively evaluates all candidate models within the search space, making it best suited for small search spaces with a limited number of dimensions. Due to its exhaustive nature, it is not practical for large search spaces with millions of possible models.
"GA" (Genetic Algorithm): The genetic algorithm is an evolutionary optimization technique inspired by natural selection and genetics. It employs techniques such as selection, crossover, and mutation to evolve a population of candidate models over multiple generations. By applying natural selection principles, the genetic algorithm aims to converge towards better-performing models. It is suitable for moderate to large search spaces and can handle a diverse range of problem types.
"GP" (Gaussian Process Algorithm): The Gaussian Process is one of the two options used in Bayesian Optimization. It specifies the form of the prior and posterior distribution for model evaluations. Initially, the distribution is random, similar to other global search algorithms. As models are executed and their results are obtained, the distribution is updated using the "ask" and "tell" steps. The Gaussian Process aims to use probabilistic models to guide the search towards promising regions of the search space efficiently. It is particularly useful for expensive-to-evaluate functions and can handle both continuous and discrete variables.
"RF" (Random Forest Algorithm): The Random Forest algorithm is an ensemble learning method that constructs multiple decision trees during the optimization process. It leverages bagging and random feature selection to increase the precision of tree building. By combining multiple trees, the Random Forest aims to achieve higher accuracy and robustness in the optimization process. It is effective for a wide range of problem types and can handle both regression and classification tasks.
"GBRT" (Gradient Boosted Random Tree Algorithm): The Gradient Boosted Random Tree algorithm is a variation of the Random Forest approach. It builds trees progressively by calculating the gradient of the reward or fitness with respect to each decision. This allows the algorithm to focus on challenging regions of the search space, which can lead to increased precision and improved optimization results. Similar to Random Forest, it is suitable for regression and classification problems.
"PSO" (Particle Swarm Optimization Algorithm): The Particle Swarm Optimization algorithm is a population-based optimization technique inspired by the social behavior of bird flocks or fish schools. It represents potential solutions as particles that move through the search space to find the optimal solution. Particles communicate and share information about their current best-known positions, allowing them to explore promising areas collaboratively. The Particle Swarm Optimization is effective for continuous optimization problems and can handle noisy or multimodal objective functions.
When using the create_pyDarwinOptions function, you can specify one of these algorithm values to choose the appropriate optimization strategy for your specific problem. Each algorithm has its strengths and limitations, and the choice of algorithm should be based on the problem's characteristics and the desired search space exploration behavior.
Please see pyDarwin documentation for more details.#'
# Create pyDarwin options with default values
pyDarwinOptions <- create_pyDarwinOptions()
# Create pyDarwin options with custom author and algorithm
pyDarwinOptions <-
create_pyDarwinOptions(author = "John Doe",
algorithm = "PSO")
Run the code above in your browser using DataLab