Machine learning scripts
The ML library in Pyramid 2018 is a very nice addition. However, is it possible to see the scripts behind the algorithms (as was the case in BI Office)? Now the algo allows to tweak the arguments in a very limited manner and it is not possible to know what assumptions (e.g. node rules for decision trees and number of trees in random forest) are used in the algo logic and it makes interpretation and improving of the model difficult. Or is it assumed that deeper refining of the algos should be done by importing the script by the user as R or Python code?
Br,
Simo
2 replies
-
Hey Simo,
The ML nodes are Java\Scala implementation by Weka, Mllib and DL4J, the different implementation supports different arguments. Currently you can’t change those via pyramid, we will add this down the road.
If you’d like to use scripts, you can either use the marketplace that supports both R and Python (which you can tweak) or you can create your own script as save it for future use.
For more information about the Java implementation:
Weka: https://www.cs.waikato.ac.nz/ml/weka/
Mllib: https://spark.apache.org/docs/2.2.0/ml-guide.html
DL4J: https://deeplearning4j.org/
Hope that helps,
Bar
-
Hi Bar,
Thanks, this clarifies.
Br,
Simo