Bringing back uncertainty to machine learning.
Boot.fit
and Boot.predict
methods are now parallelised, speeding up both training
and prediction time a bit.README
to include generalised linear models, rather than only
mentioning linear regression.PyTorch
model support, as that has not been implemented
yetverbose
argument to QuantileRegressionForest
, which displays a
progress bar during training.QuantileRegressionForest.min_samples_leaf
has changed
from 1 to 5, to ensure that the quantiles can always be computed sensibly
with the default setting.logkow
feature in the FishBioconcentration
dataset is now converted
into a float, rather than a string.README
quantiles
argument to QuantileRegressionTree
and Boot
, as an
alternative to specifying uncertainty
, if you want to return specific
quantiles.QuantileRegressor
, which can wrap any general linear model
for quantile predictions.Boot.predict
were based on a fitting of the model to one
of the bootstrapped datasets. It is now based on the entire dataset, which in
particular means that the predictions will be deterministic. The intervals
will still be stochastic, as they should be.Boot
are now calculated during fitting, which should
decrease the prediction times a tiny bit.statsmodels
score
method to QuantileLinearRegression
, which either
outputs the mean negative pinball loss function, or the R^2 valueQuantileLinearRegression
QuantileLinearRegression
Boot.predict
, where the definition of generalisation
should be the
difference of the means of the residuals, and not the difference between
the individual quantiles. Makes a very tiny difference to the prediction
intervals. Thanks to Bryan Shalloway for catching this mistake.