Select the Model selection algorithm link in the left list.
Select the search Algorithm from the pulldown.
GA: Genetic Algorithm that reproduces the mathematics of evolution/survival of the fittest. If selected, a GA setup section and Downhill setup section are added containing algorithm-specific options.
GP: Gaussian Process that involves specifying the form of the prior and posterior distribution. If selected, a Downhill setup section is added.
RF: Random Forest consists of splitting the search space into “good” and “bad” regions. If selected, a Downhill setup section is added.
GBRT: Gradient Boosted Random Tree is similar to RF but progressively builds the tree, calculating a gradient of reward/fitness with each decision. If selected, a Downhill setup section is added.
EX: Exhaustive search, in which the search space is initially represented as a string of integers - one for each dimension.
PSO: Particle Swarm Optimization algorithm, like GA, attempts to reproduce a natural optimization process. In the case of PSO, the natural process is the swarm behavior of birds and fish. Each candidate NONMEM model moves through the search space, as one might imagine individuals in a school of fish or a flock of birds moving together but also somewhat independently to search for global optimal solution in the candidate model search space. If selected, a PSO setup section and Downhill setup section are added containing algorithm-specific options.
In the Number of generations field, set the number of iterations of the search algorithm to run (not used by EX algorithm). The value must be a positive number.
In the Number of models in every generation field, set the number of models to create in each generation. This is the population of each generation (not used by EX algorithm). The value must be a positive number.
In the Size of exhaustive search model batch field, enter the number of models to run as a batch. (Only applies to the EX algorithm.) The value must be a positive number.
In the Number of parallel processes to perform the “ask” step field, enter the number of processes to run in parallel when generating samples for the next generation (not used by GA, PSO, or EX algorithms). The value must be a positive number.
In the Seed value for random number generator field, enter a positive integer to serve as the seed for the random number generator (not used by EX algorithm).