The Active Learning Loop is an extension to the regular Recursive Loop with an additional model port. It enables passing of a data table and an arbitrary model from the Active Learning Loop End back to the Active Learning Loop Start.
The Loop Start requires an initialized table and model, which will be output by the Active Learning Loop Start in the first iteration of the loop.
The table and model received by the corresponding Active Learning Loop End is passed back to the Active Learning Loop Start node. Starting with the second iteration, the Active Learning Loop Start node outputs the table and model as received by the Active Learning Loop End.
The loop runs until one of the three stopping criteria is met:
- Maximum number of iterations: to ensure no endless loop is created, the loop will end after the set number of iterations.
- Minimal number of rows: to ensure enough rows are present for processing, the loop stops if its input contains less rows than the set minimum.
- End loop with variable: the loop ends if the option is enabled and the value of the selected variable equals "true"
The recursion model output (1) will always output the model from the last iteration connected to the recursion model input (1). While the loop is running, the model will be returned to the Active Learning Loop Start node. The table passed to the collecting table input (2) is collected and passed to the collected table output port (2). The table passed to the recursion table input (3) is returned to the Recursive Loop Start node during the iteration.