At this line, it may be better to explicitly mention which parameters will be sampled with replacement if any one of them is a distribution:
https://github.com/databricks/spark-sklearn/blob/master/python/spark_sklearn/random_search.py#L27
Are all parameters (those given as distributions and those given as lists of values) sampled with replacement in this scenario, or only parameters given by distributions?
If so, this documentation would be better stated as:
sampling with replacement is used for all parameters.