ternarynet (version 1.16.0)

ternaryFitParameters-class: Ternary Network Fitting Parameters

Description

This is a class representation of the input parameters for the ternary network fitting algorithm implemented in the function tnetfit.

Creating Objects

ternaryFitParameters() This creates a ternaryFitParameters object with the default fitting parameters.

Slots

perturbationType:

this parameter currently can only be set to 1

scoreType:

the method to score networks. Can be set to either 1 or 2, corresponding the the score types in Almudevar et al. (2011).

backupStage:

current fit is output periodically according to this parameter

maxStage:

the maximum number of stages permitted. Ideally, the actual number of stages required until convergence should be much less than this value.

maxTransition:

This parameter provides an adaptive truncation of the stage sample size. The stage terminates before the specified fixed sample size if the number of transitions resulting in a strict increase of the score reaches this value. If the sampler is in steady state, then this count should be approximately half the number of transitions in which the score changes value.

epsilon:

Convergence tolerance.

beta0:

Algorithm terminates when this number of consecutive convergence events have occurred.

chi0:

The target initial acceptance rate. This should be close to 1, although setting it too close will increase computation time.

delta:

The increment change in steady state distribution between stages (as variational distance). Larger values tend to decrease computation time, but too large a value will result in spurious convergence.

ne:

The fixed sample size (number of MCMC transitions) per stage.

m0:

The sample size (number of transitions) used to determine the initial temperature.

maxDegree:

Maximum number of parents per node permitted in model topology.

pAddParent:

This is the probability of adding a parent to a randomly selected node in the proposal function.

pExchangeParent:

This parameter gives the probability of a parent exchange in the proposal function. The AddParent operation takes precedence, so this probability should be interpreted as being conditional on the rejection of the AddParent operation.

neighborDegree:

Number of applications of the proposal function.

pNeighborhood:

Vector of probabilities denoted, which generates the random number of proposal function iterations. The length is one less than neighborDegree. If neighborDegree equals 1 then no iteration is performed, and this vector is ignored.

rho:

Weight parameter for the exponential smoothing of the variance estimate. For no smoothing set to 1.

edgePenalty:

This parameter provides a complexity penalty. This number times the number of edges is added to the score. To apply no penalty set this parameter to 0.

Methods

All named elements can be accessed and set in the standard way (e.g. scoreType(object) and scoreType(object)<-).

See Also

tnetfit, ternaryFit-class, ternaryPost-class. Almudevar A, McCall MN, McMurray H, Land H (2011). Fitting Boolean Networks from Steady State Perturbation Data, Statistical Applications in Genetics and Molecular Biology, 10(1): Article 47.

Examples

# create an instance
ternaryFitParameters()