Should min_child_weight be set based on the number of instances in the overall dataset, or the number of instances in each query ID?
Should min_child_weight be set based on the number of instances in the overall dataset, or the number of instances in each query ID?