CliMA / CliMA/CalibrateEmulateSample.jl

RF localisation causes singularity issues

Open
#288 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Julia
Stars
90
Forks
16
PR merge metrics
No merged PRs in 30d

Description

## issue
During optimization of RF hyperparameters, one of the loss function entries was zero-d out (a hack fix to remove this item).
It does not outwardly effect the calibration. But this causes ensembles in the optimisation algorithm to have a 0-uncertainty row, which causes an issue with the localisation algorithms that often invert the variance/covariance.

## Solution

Remove the item from the loss function&data instead of simply zeroing it out.

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by tracing where RF hyperparameter optimization constructs the loss function and data, then follow the localisation code that inverts variance or covariance. The issue is done when the affected item is removed from the loss function and data rather than zeroed, so ensembles no longer contain a zero-uncertainty row.

Written by the indexing model from the issue text.

Assessment

Tech stack
julia
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.