[Bug] Bayesian Approach #28

Closed
opened 3 years ago by AlessandroBregoli · 0 comments
AlessandroBregoli commented 3 years ago (Migrated from github.com)
Owner

Basic info

  • Library Version: 0.1.0
  • Branch/commit: dev/2d7e52f8f1f2e4594c69285b157ebe744d04e117

Expected Behavior

The Bayesian approach (parameter learning) divides alpha and tau by the number of possible configurations in its parent set.

Actual Behavior

The library use alpha and tau ignoring the parent set.

Steps to Reproduce

Irrelevant.

Log tracebacks and/or errors shown

Irrelevant.

(Optional) Extra info

Without recomputing the alpha with respect to parent set of the node, the equivalent sample size defined by the prior distribution increases with the number of parents. When there are few data this leads to ignoring the dataset in favor of the prior distribution.

## Basic info * **Library Version:** 0.1.0 * **Branch/commit:** dev/2d7e52f8f1f2e4594c69285b157ebe744d04e117 ## Expected Behavior The Bayesian approach (parameter learning) divides alpha and tau by the number of possible configurations in its parent set. ## Actual Behavior The library use alpha and tau ignoring the parent set. ## Steps to Reproduce Irrelevant. ## Log tracebacks and/or errors shown Irrelevant. ## (Optional) Extra info Without recomputing the alpha with respect to parent set of the node, the equivalent sample size defined by the prior distribution increases with the number of parents. When there are few data this leads to ignoring the dataset in favor of the prior distribution.
Sign in to join this conversation.
No Milestone
No Assignees
1 Participants
Notifications
Due Date

No due date set.

Dependencies

No dependencies set.

Reference: unimib-thesis-mad/reCTBN#28
Loading…
There is no content yet.