VariantSpark
VariantSpark copied to clipboard
aehrc
machine learning for genomic variants
If possible for each batch of trees (or better if for each tree) the training parameter (below) is read from a file and apply to the training. - batch (or...
If the model file can have the following metadata for - RF - each batch of tree - each tree - each node where applicable. Metadata: generally all parameter used...
similar to the stdout in CLI if the hail interface prints a message after building each batch of trees.
If it could be possible to merge multiple Java model file into one and then perform any of these tasks - Importance-Score computation - Prediction - Conversion to JSON or...
If there would be a command to convert Java RandomForest models to the model supported by R Random Forest. This helps to apply post-processing of RF model with existing algorithm...
If there would be a command to convert Java and JSON models.
If this could be possible to separate the training process from the importance-score computation. The training process may optionally perform importance-score computation too. The importance-score computation accept one or multiple...
That would be great if VariantSpark output the RandomForest model continuously. For example, after building a batch of trees, save trees in a separate model file (Java or Json) In...
implement mTry as outlined in` RandomForestArgs.scala` `@Option(name="-rmt", required=false, usage="RandomForest: mTry(def=sqrt())" , aliases=Array("--rf-mtry")) val rfMTry:Long = -1L ` To be -accessible in cli and api modules
