Models

Index

• UncertaintyQuantification.LinearBasisFunctionModel
• UncertaintyQuantification.Model
• UncertaintyQuantification.ParallelModel
• UncertaintyQuantification.UQModel
• UncertaintyQuantification.evaluate!
• UncertaintyQuantification.evaluate!
• UncertaintyQuantification.reliability

Types

UncertaintyQuantification.UQModel Type

Abstract supertype for all model types

source

UncertaintyQuantification.Model Type

Model(f::Function, name::Symbol)

The function f must accept a DataFrame and return the result of the model for each row in the DataFrame as a vector. The name is used to add the output to the DataFrame.

source

UncertaintyQuantification.ParallelModel Type

ParallelModel(f::Function, name::Symbol)

The ParallelModel does what the Model does with a small difference. The function f is passed a DataFrameRow not the full DataFrame. If workers (through Distributed) are present, the rows are evaluated in parallel.

source

UncertaintyQuantification.LinearBasisFunctionModel Type

LinearBasisFunctionModel(
    df::DataFrame, out::Symbol, b:<AbstractBasis, inputs::Vector{Symbol}=propertynames(df[:, Not(out)])
)

Construct a linear basis function model for the data in df using the basis b.By default the input variables are assumed to be all columns of the DataFrame except for out.

source

Methods

UncertaintyQuantification.evaluate! Method

evaluate!(m::Model, df::DataFrame)

Calls m.func with df and adds the result to the DataFrame as a column m.name

source

UncertaintyQuantification.evaluate! Method

evaluate!(m::ParallelModel, df::DataFrame)

Calls m.func for each row of df and adds the result to the DataFrame as a column m.name. If workers are added through Distributed, the rows will be evaluated in parallel.

source

UncertaintyQuantification.reliability Method

reliability(ipm::IntervalPredictorModel, ϵ::Real)

Returns the confidence parameter $\beta \in (0,1)$, such that the reliability of the IPM, that is the probability unobserved data points will fall in its bounds, is no less than $1-ϵ$ with confidence $1-\beta$, with $ϵ\in (0,1)$.

source