optimizeByLOESS function

Function to optimize a two-part strategy breeding scheme: 1. Simulate a batch using given percentage ranges 2. Perform LOESS fit to the gains 3. Find budget with best estimated gain 4. Calculate new percentage ranges: any simulation within 2*StdErr of best 5. Decide on some simulations to repeat: 1. Parameter space with high gain and high std err: need more info there 2. Parameter space with high gain: high probability that it's best Go back to 1.

optimizeByLOESS(
  batchSize,
  targetBudget,
  percentRanges,
  startCycle,
  wgtPopImprov,
  tolerance,
  baseDir = NULL,
  maxNumBatches = 10,
  initializeFunc,
  productPipeline,
  populationImprovement,
  bsp,
  randomSeed = NULL,
  nCores = 1
)

Arguments

batchSize

Integer number of simulations between LOESS fits
targetBudget

Numeric value that you want the budget adjusted to
percentRanges

Numeric matrix (nStages + 1 rows and 2 columns) of percentage budget allocation to crossing (F1) and all of the stages. If there is a stage that is genotyped, the genotyping cost is added to that stage
startCycle

Integer the start cycle from which to measure gain. The end cycle will be the last cycle. Set the startCycle so there is enough burn-in
wgtPopImprov

Two components contribute to the utility: yearly population improvement and yearly gain in the product development. This is the weight for the first. The other has weight 1 - wgtPopImprov
tolerance

Numerical difference between min amd max percentage budgets for all stages
baseDir

Directory if you want to have progress saved by batch. Relative to R working directory. If not empty string, include final /
maxNumBatches

Integer to stop the simulations eventually if the algorithm is not narrowing in on optimal parameter values
initializeFunc

Function for runBreedingScheme
productPipeline

Function for runBreedingScheme
populationImprovement

Function for runBreedingScheme
bsp

A list of breeding scheme parameters.
randomSeed

Integer seed for random number generator
nCores

Integer number of cores to use for parallel simulation

Value

Numeric matix with all simulations budget allocations, gen mean change, gen std dev change, total cost.

Details

A wrapper to repeatedly simulate a scheme with different budget allocations to find optimal allocations

Examples