Optimus: Acceptance Ratio Simulated Annealing and Acceptance Ratio Replica Exchange Monte Carlo Optimisation Engine

Description

Acceptance Ratio Simulated Annealing and Acceptance Ratio Replica Exchange Monte Carlo Optimisation Engine

Usage

Optimus(
  NUMITER = 1e+06,
  STATWINDOW = 70,
  T.INI = 1e-05,
  T.ADJSTEP = 5e-09,
  TSCLnum = 2,
  T.SCALING = 3,
  T.MIN = 5e-09,
  T.DELTA = 2,
  DUMP.FREQ = 10000,
  LIVEPLOT = TRUE,
  LIVEPLOT.FREQ = 1e+05,
  PDFheight = 29,
  PDFwidth = 20,
  NCPU = 4,
  LONG = TRUE,
  SEED = 840,
  OPTNAME = "",
  DATA = NULL,
  K.INITIAL = 0,
  rDEF,
  mDEF,
  uDEF,
  EXCHANGE.FREQ = 1000,
  ACCRATIO = c(90, 50, 5, 1),
  CYCLES = 10,
  ACCRATIO.IN = 90,
  ACCRATIO.FIN = 0.5,
  OPT.TYPE = "SA",
  DIR,
  starcore = NULL
)

Value

A probabilistic optimal parameter configuration K.

Arguments

NUMITER: Number of model optimisation steps.
STATWINDOW: Number of last ongoing iterations to calculate acceptance ratio for temperature auto-adjustment.
T.INI: Initial temperature (K) for Metropolis criterion.
T.ADJSTEP: Temperature change step-size for temperature auto-adjustment based on the actual acceptance ratio.
TSCLnum: Cutoff for one of the NumofAccRatSMIdeal and NumofAccRatGRIdeal numbers after which the adjustment step is multiplied by T.SCALING.
T.SCALING: See above.
T.MIN: Value to which the pseudo-temperature is set when the
T.DELTA: Minimum value by which acceptance ratio in a STATWINDOW must differ from the ideal acceptance ratio for the temperature control unit to make a temperature adjustment.
DUMP.FREQ: The frequency (in steps) of writing the found ongoing best model.
LIVEPLOT: Plotting the optimisation process in a pdf file.
LIVEPLOT.FREQ: Frequency (in steps) of plotting the results.
PDFheight: Plot height in inches.
PDFwidth: Plot width in inches.
NCPU: Number of CPU cores to use, by running more replicas of the optimisation. The usage of more than 1 cores will attempt to load the foreach and doParallel libraries in the case of SA Optimus
LONG: If TRUE, it means that a long simulation is expected to be done, hence the memory-friendly mode will be activated.
SEED: Setting the seed for the random number generator.
OPTNAME: The name of the optimisation process.
DATA: A list that holds any supplementary data that functions mDEF or uDEF need to access.
K.INITIAL: The initial parameter configuration from which the optimisation process will begin.
rDEF: Function that defines a rule by which the parameters(K) are randomly altered.
mDEF: Model function that operates on the parameters to be optimized (K) and returns an observable object O.
uDEF: Function that evaluates the performance of a given set of parameters K.
EXCHANGE.FREQ: Frequency of exchanges (NUMITER should be divisible by this number, for RE Optimus).
ACCRATIO: Vector of Acceptance Ratios for each replica (length of ACCRATIO must be equal to NCPU, for RE Optimus).
CYCLES: Number of annealing cycles (NUMITER should be divisible by this number, for SA Optimus).
ACCRATIO.IN: Initial acceptance ratio (%) at the beginning of each constituent annealing cycles (for SA Optimus).
ACCRATIO.FIN: Final acceptance ratio (%) at the end of each constituent annealing cycles (for SA Optimus).
OPT.TYPE: String specifying which optimisation protocol to use. Enter "SA" for Simulated Annealing or "RE" for Replica Exchange (default value is "SA")
DIR: String specifying which optimisation protocol to use.
starcore: Experimental variable of type list, holding some parameters for in-lab starcore use only.

Examples

Run this code


K <- IJ_ORIG
K$j <- sample(x=K$j, size=nrow(K), replace=FALSE)
out.dir <- tempdir()

Optimus(NCPU=1, OPTNAME="IJ.NEW.OPTI.SA", NUMITER=500, CYCLES=2, DIR=out.dir,
        DUMP.FREQ=10, LONG=FALSE, OPT.TYPE="SA", K.INITIAL=K,
        rDEF=ex.r.fun, mDEF=ex.m.fun, uDEF=ex.u.fun,
        DATA=list(IJ_ORIG=IJ_ORIG, gaplimit=50, numContacts=nrow(IJ_ORIG)))

Run the code above in your browser using DataLab