How can I understand the working of regression cases (continuous variables)?

[CountingLogic]

As far as I understand, Boost SRL allows for only one query predicate. Now, in the regression case you discretise the predicates, which is clear for the non-query predicates, but once you discretise the query predicates, then you have many more queries, how is this handled ? Also, If I would like to interpret the decision trees as FOL rules, how can I do it in this case?

[boost-starai]

Thanks for the interest. The issue is that we do NOT discretize the target variable. So if you prefer to do regression, the gradients are now simply v-v’ where v is the true value and v’ is the current predicted value. This can again be found in several of our papers. This gradient is used for fitting the tree. Again, we do NOT discretize the target. Only the non-query predicates are discretized. Best, SN From: SagarMal ***@***.*** Sent: Sunday, May 30, 2021 5:54 AM To: starling-lab/BoostSRL ***@***.***> Cc: Subscribed ***@***.***> Subject: [External] [starling-lab/BoostSRL] How can I understand the working of regression cases (continuous variables)? (#42) This message was sent from a non-IU address. Please exercise caution when clicking links or opening attachments from external sources. As far as I understand, Boost SRL allows for only one query predicate. Now, in the regression case you discretise the predicates, which is clear for the non-query predicates, but once you discretise the query predicates, then you have many more queries, how is this handled ? Also, If I would like to interpret the decision trees as FOL rules, how can I do it in this case? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#42>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AF4KTVO6MXHYZWGKRCPDDETTQIKNNANCNFSM45ZHJMFA>.