adds support to sparse matrix input in the form of dgCMatrix
solitude 1.1.1
fit gives a message when input dataset has duplicate and does not
throw an error
solitude 1.1.0
fit exposes max_depth parameter with default value of
ceiling(log_2(sample_size))
path length extension is not done anymore (Thanks to
karthi1870)
ranger’s forest object is now public
fit checks for duplicates in the input dataset
solitude 1.0.1
finalized on the API
isolationForest object no longer calculates and stores anomaly
scores on training data. Use predict method for this.
mtry is no longer exposed
version 1.0.0 was a faulty release, do not use it
solitude 0.2.1
added ‘mtry’ argument for isolationForest method
minor optimizations in ‘terminalNodesDepth’ function
solitude 0.2.0
Changes in API: isolationForest is now a R6 class with
new(), fit() and predict() methods.
Exported terminalNodesDepth function to obtain the
depth of terminal nodes of all trees in a ranger forest.
solitude 0.1.3
Changed defaults to avoid bagging in extremely randomized trees
(see
https://stats.stackexchange.com/questions/406666/what-exactly-is-the-extratrees-option-in-ranger)