@dariasor dariasor released this Feb 11, 2017

Assets 4
  1. Old logs are not deleted any more during the run of ag_train or bt_train. Instead they are appended to logs.archive.txt
  2. bt_train now generates a new output file: prediction scores on the validation data set. Default file name is preds.txt
  3. For ag_train, a warning appears when there is no signal (even the simplest model has performance worse than baseline).
  4. Changed the output for the layered mode of training for Additive Groves. The best model parameters and the expanding recommendation now correspond to the best model for interaction detection, not to the model providing the best performance on the validation set.
  5. File performance.txt (Additive Groves output) now contains binary information about the learning curve convergence. 1 means the learning curve has converged for given (a, n) parameters, 0 means it has not.
2.4

@dariasor dariasor released this Mar 23, 2016

Assets 4
  1. Feature evaluation algorithm in bagged trees is changed. Now a score in each node is normalized by the entropy of the split feature in that node. This way the scores of binary features become comparable with the scores of continuous features with multiple values.
  2. Effect and interaction plots now consider all data, including data points with missing values. For the features with substantial number of missing values, the effect of missing value is also plotted.