May I know if we can use loop nodes to stack operations on given data table?
condition | operator | num_to_operate
band1 | "+" | 300
band1 | "*" | 2
band2 | "+" | 1000
input data table:
name | band | salary
lily | band1 | 5000
lucy | band2 | 4569
nancy | band1 | 3850
then, I'd like to operate according to operations table from top to bottom, in order to get final salary according to base salary. Here is the result I expect:
Output data table:
name | band | salary
lily | band1 | 12600
lucy | band2 | 5569
nancy | band1 | 8300
I know I can use "TableRow To Variable Loop Start" node to start a loop, then manipulate data according to variable. But when I get pass loop end and go back loop start node, how can I use datatable which has just been manipulated in input data table, instead of the very original table?
Currently, I just save the data to hard disk before loop end, and when next loop starts, I read the table again from hard disk......How can I do it in a more elegant way?
Uuuuh I'm thinking.
I think the delegating loops (in the ensemble plugin) would help here....
My main problem right now.. How would I end it...
Because it actually ends when there are no more data points or after an fixed number of iterations.....
Ok... I made you an Example. It mainly demonstrates how the feedback with the delegating works in this case....
Hope that helps,
Thanks a lot!
"The data in port 1 is returned to the delegating loop start." -- this is the exact function I would need.
May I know why it is not under "flow control" category and what is the meaning of "delegating"?
I totally agree, I made this comment in a previous post (3rd bullet point).
No-one knows this functionality exists under its current name and location.
the term delegating stems from ensemble learning. The idea of delegating learning, is to in each run learn everything, and afterwards examples being not correctly classified are delegated to the next learner.
Here is a paper about it http://dl.acm.org/citation.cfm?id=1015395
When we developed the node we were not aware there would be so many other use cases of this functionality.
Em, data mining would be too narrow a positioning of this great tool. I did understand somewhat about data mining, which prevented me from using this great tool in the very past.
It was very recent that I was searching for open source alternative for Brio/Hyperion and ran into this list(http://www.kdnuggets.com/polls/2010/data-mining-analytics-tools.html). After I tried tools in the list, I then understand that it could serve in much big scope!