Row-wise tensor product base learner — BaselearnerTensor • compboost

This class combines base learners. The base learner is defined by a data matrix calculated as row-wise tensor product of the two data matrices given in the base learners to combine.

Format

S4 object.

Arguments

blearner1: (Baselearner*)
First base learner.
blearner2: (Baselearner*)
Second base learner.
blearner_type: (character(1))
Type of the base learner (if not specified, blearner_type = "spline" is used). The unique id of the base learner is defined by appending blearner_type to the feature name: paste0(blearner1$getDataSource()getIdentifier(), "_", blearner2$getDataSource()getIdentifier(), "_", blearner_type).
anisotrop: (logical(1))
Defines how the penalty is added up. If anisotrop = TRUE, the marginal effects of the are penalized as defined in the underlying factories. If anisotrop = FALSE, an isotropic penalty is used, which means that both directions gets penalized equally.

Usage


BaselearnerTensor$new(blearner1, blearner2, blearner_type)
BaselearnerTensor$new(blearner1, blearner2, blearner_type, anisotrop)

Fields

This class doesn't contain public fields.

Methods

$summarizeFactory(): () -> ()
$transfromData(newdata): list(InMemoryData) -> matrix()
$getMeta(): () -> list()

Inherited methods from Baselearner

$getData(): () -> matrix()
$getDF(): () -> integer()
$getPenalty(): () -> numeric()
$getPenaltyMat(): () -> matrix()
$getFeatureName(): () -> character()
$getModelName(): () -> character()
$getBaselearnerId(): () -> character()

Examples

# Sample data:
x1 = runif(100, 0, 10)
x2 = runif(100, 0, 10)
y = sin(x1) * cos(x2) + rnorm(100, 0, 0.2)
dat = data.frame(x1, x2, y)

# S4 wrapper

# Create new data object, a matrix is required as input:
ds1 = InMemoryData$new(cbind(x1), "x1")
ds2 = InMemoryData$new(cbind(x2), "x2")

# Create new linear base learner factory:
bl1 = BaselearnerPSpline$new(ds1, "sp", list(n_knots = 10, df = 5))
bl2 = BaselearnerPSpline$new(ds2, "sp", list(n_knots = 10, df = 5))

tensor = BaselearnerTensor$new(bl1, bl2, "row_tensor")

# Get the transformed data:
dim(tensor$getData())
#> [1] 196 100

# Get full meta data such as penalty term or matrix as well as knots:
str(tensor$getMeta())
#> List of 3
#>  $ df         : num [1:2, 1] 5 5
#>  $ penalty    : num [1:2, 1] 15.3 15.6
#>  $ penalty_mat: num [1:196, 1:196] 30.9 -31.1 15.6 0 0 ...

# Transform "new data":
newdata = list(InMemoryData$new(cbind(runif(5)), "x1"),
  InMemoryData$new(cbind(runif(5)), "x2"))
str(tensor$transformData(newdata))
#> List of 1
#>  $ design:Formal class 'dgCMatrix' [package "Matrix"] with 6 slots
#>   .. ..@ i       : int [1:80] 0 2 4 0 1 2 3 4 0 1 ...
#>   .. ..@ p       : int [1:197] 0 3 8 13 18 20 20 20 20 20 ...
#>   .. ..@ Dim     : int [1:2] 5 196
#>   .. ..@ Dimnames:List of 2
#>   .. .. ..$ : NULL
#>   .. .. ..$ : NULL
#>   .. ..@ x       : num [1:80] 3.33e-03 1.67e-07 5.85e-04 5.73e-02 2.74e-03 ...
#>   .. ..@ factors : list()

# R6 wrapper

cboost = Compboost$new(dat, "y")
cboost$addTensor("x1", "x2", df = 5)
cboost$train(50, 0)
#> Train 50 iterations in 0 Seconds.
#> Final risk based on the train set: 0.12
#> 

table(cboost$getSelectedBaselearner())
#> 
#> x1_x2_tensor 
#>           50 
plotTensor(cboost, "x1_x2_tensor")