presample: Forest-wide Observation Sampling

In Rborist: Extensible, Parallelizable Implementation of the Random Forest Algorithm

Forest-wide Observation Sampling

Description

Observations sampled for each tree to be trained. In the case of the Random Forest algorithm, this is the bag.

Usage

## Default S3 method:
presample(y,
                            samplingWeight = numeric(0),
                            nSamp = 0,
                            nRep = 500,
                            withRepl =  TRUE,
                            nHoldout = 0,
                            nFold = 1,
                            verbose = FALSE,
                            nTree = 0,
                            ...)

Arguments

y	A vector to be sampled, typically the response.
samplingWeight	Per-observation sampling weights. Default is uniform.
nSamp	Size of sample draw. Default draws y length.
nRep	Number of samples to draw. Replaces deprecated nTree.
withRepl	true iff sampling is with replacement.
nHoldout	Number of observations to omit from sampling. Augmented by unobserved response values.
nFold	Number of collections into which to partition the respone.
verbose	true iff tracing execution.
nTree	Number of samples to draw. Deprecated.
...	not currently used.

Value

an object of class Sampler consisting of:

• yTrain the sampled vector.

• nSamp the sample sizes drawn.

• nRep the number of independent samples.

• nTree synonymous with nRep. Deprecated.

• samples a packed data structure encoding the observation index and corresponding sample count.

• hash a hashed digest of the data items.

References

Tille, Yves. Sampling algorithms. Springer New York, 2006.

Examples

  ## Not run: 
    y <- runif(1000)

    # Samples with replacement, 500 vectors of length 1000:
    ps <- presample(y)

    # Samples, as above, with 63 observations held out:
    ps <- presample(y, nHoldout = 63)

    # Samples without replacement, 250 vectors of length 500:
    ps2 <- presample(y, nTree=250, nSamp=500, withRepl = FALSE)


  
## End(Not run)

Rborist documentation built on April 3, 2025, 8:04 p.m.