This method should be implemented with speed and garbage collection in mind, as it called once per row on potentially huge data-sets.
This method should be implemented with speed and garbage collection in mind, as it called once per row on potentially huge data-sets.
This method is not required to be thread-safe.
The row of input to be scored.
The result from applying the trained model to the row.
Allows the transformer to specify if the apply method can handle null values in the input row.
Allows the transformer to specify if the apply method can handle null values in the input row.
e.g. a Null value replacement transformer, or a Naive Bayes transformer would naturally handle null values, but a Linear Regression transformer would not.
Default value is false.
Boolean indicating tolerance for null values in input.
Serialization doesn't have to be maintained between different versions for this object, since it it always created by a model, but it does need to be serializable for use in Spark jobs (for sending to worker nodes).