You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Compute sign inside SignedProbabilities and replace 0 probabilities
Since random forest may produce a probability of 0, we replace such
values with 1e-100 to retain the sign (positivity vs negativity of LP
eddges).
Adding the sign as an argument to SignedProbabilities::add makes it
impossible to add a 0 value with ambiguos sign.
If the logic was at the call-site, then future call-sites would be
unprotected from such faults and would have to know about the 1e-100
constant that replaces 0.0.
Co-Authored-By: Mats Rydberg <[email protected]>
We can see the model configuration with `tolerance = 0.001` (and defaults filled for remaining parameters) was selected, and has a score of `0.76` on the test set.
240
+
We can see the RandomForest model configuration with `numberOfDecisionTrees = 5` (and defaults filled for remaining parameters) was selected, and has a score of `0.58` on the test set.
241
241
The score computed as the <<linkprediction-pipelines-metrics, AUCPR>> metric, which is in the range [0, 1].
242
242
A model which gives higher score to all links than non-links will have a score of 1.0, and a model that assigns random scores will on average have a score of 0.5.
Copy file name to clipboardExpand all lines: pipeline/src/main/java/org/neo4j/gds/ml/pipeline/linkPipeline/train/LinkPredictionEvaluationMetricComputer.java
0 commit comments