simulate_extra
Simulate extrapolated data
Simulated data scenarios described in the paper from Ryan and Culp (2015).
Usage
simulate_extra(n_source = 100, n_target = 100, p = 1000, shift = 10,
scenario = "same", response = "linear", sigma2 = 2.5)
Arguments
- n_source
Number of source samples (labeled)
- n_target
Number of target samples (unlabeled)
- p
Number of variables (
p > 10
)- shift
The shift applied to the first 10 columns of xU.
- scenario
Simulation scenario. One of
"same"
(same distribution),"lucky"
(extrapolation with lucky \(\beta\)),"unlucky"
(extrapolation with unlucky \(\beta\))- response
Type of response:
"linear"
or"logit"
- sigma2
The variance of the error term, linear response case.
Value
A list, with
- xL
data frame with the labeled (source) data
- yL
labels associated with
xL
- xU
data frame with the unlabeled (target) data
- yU
labels associated with
xU
(for validation/testing)
References
Ryan, K. J., & Culp, M. V. (2015). On semi-supervised linear regression in covariate shift problems. The Journal of Machine Learning Research, 16(1), 3183-3217.
See Also
Examples
# NOT RUN {
set.seed(0)
data = simulate_extra()
train = s2Data(data$xL, data$yL, data$xU)
valid = s2Data(data$xU, data$yU, preprocess = train)
model = s2netR(train, s2Params(0.1))
ypred = predict(model, valid$xL)
plot(ypred, valid$yL)
# }