-
Notifications
You must be signed in to change notification settings - Fork 8
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
No test covers the empty-CAS guard in _predict_admixed
correctnessWrong or fragile behaviourWrong or fragile behaviourStatus: Open.#294 In dvitale199/GenoTools;"CAS" is hardcoded in the admixture step, but the label vocabulary is user-supplied
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationcorrectnessWrong or fragile behaviourWrong or fragile behaviourStatus: Open.#293 In dvitale199/GenoTools;Ancestry predict path holds three full-size copies; nothing bounds the peak
perfMemory, speed, parallelismMemory, speed, parallelismrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#292 In dvitale199/GenoTools;Retraining moves ~1.3% of labels: a plateaued grid search, not a race
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#291 In dvitale199/GenoTools;Persist cv_results_ as a record of model selection
refactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)reportingJSON report, logs, diagnostics, observabilityJSON report, logs, diagnostics, observabilityStatus: Open.#290 In dvitale199/GenoTools;Cache the UMAP fits — the grid search recomputes each embedding six times
perfMemory, speed, parallelismMemory, speed, parallelismrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#289 In dvitale199/GenoTools;Supervised UMAP is asymmetric between fit and transform
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#288 In dvitale199/GenoTools;Does UMAP + gblinear earn its cost over a model-free baseline?
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#287 In dvitale199/GenoTools;TrainingConfig.n_jobs derives from machine RAM and CPU count, so run cost is not reproducible
perfMemory, speed, parallelismMemory, speed, parallelismrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#286 In dvitale199/GenoTools;Fit-validation thresholds are reasoned, not calibrated
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#285 In dvitale199/GenoTools;Revisit updater=coord_descent only if XGBoost's linear updaters change
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#284 In dvitale199/GenoTools;Audit released ancestry label sets for the collapse signature
ancestry-qualityAncestry model accuracy, calibration, validationAncestry model accuracy, calibration, validationrefactor-backlogMigrated from REFACTOR.md (rounds 1-24)Migrated from REFACTOR.md (rounds 1-24)Status: Open.#283 In dvitale199/GenoTools;