Processing branches and subject overrides
Branches compare processing alternatives. Subject overrides adapt supported settings for particular recordings. They are different tools, with separate study-level plans.
Create a processing branch
The ERPERF General page selects the current branch and opens branch management. The ordinary chain is main. A named branch declares its parent and the first stage it recalculates:
| Fork stage (ICA-path number) | Branch-owned work |
|---|---|
| Standard filtering (7) | Filters, epochs and all later results |
| Epoch extraction (8) | Epochs and all later results |
| Artefact parameters (9) | Artefact assessment and averaging/results |
- Choose a parent and an appropriate fork point.
- Give the branch a distinct name and description.
- Edit the branch-owned settings, checking the banner above affected pages.
- Apply/OK, verify the selected processing range, and run a test recording.
Selecting a branch normally selects its fork through the final step. A run may not start before the selected branch owns the work.
Earlier settings/checkpoints come from the parent. The banner identifies the owner; with nested branches, different pages can have different owners. Switching back to main restores the workspace’s main settings. Branch edits remain local to Configure Study until Apply/OK; Cancel discards those edits.
Where branches live
Definitions are stored in ERPERF_processing-branches.json in the study.
Outputs are stored under:
erperf/<recording>/branches/<branch-name>/
Earlier large recordings are inherited, not copied. Nested branches may inherit through several parents while their physical folders remain siblings.
Each processed subject branch keeps execution metadata describing its settings and inherited checkpoint. If that parent checkpoint changes, reuse can stop because the child is stale. Recalculate from the fork rather than renaming files or bypassing the check.
Restore definitions from bundle folders… in branch management can recover lost plan entries from subject bundles. Review recovered settings, parent and fork before saving: incomplete or conflicting metadata may require inference. Recovering a definition does not recreate missing data. Inferred parent relationships default to main; restore any nested lineage explicitly rather than assuming that folder nesting recorded it.
Example: alternate epochs
A branch beginning at epoch extraction can keep the accepted ICA and standard filters but use a different epoch window. It then needs fresh artefact assessment and averages. A branch beginning at standard filtering can also change those filters. These are examples of workflow isolation, not recommended scientific parameter choices.
Subject overrides
Use Subject overrides… in the Batch Manager to create named profiles and assign them to recordings. The currently available override scope is ERPERF ICA training; other listed scopes may be marked unavailable.
The effective setup starts from the current workspace and branch settings, then applies the assigned supported subject profile. The worker receives a frozen copy and performs ordinary compatibility checks. The log identifies the profile and changed values.
Overrides are stored in ERPERF_subject-overrides.json, independently of the
workspace. Moving a study can resolve assignments by a unique filename; ambiguous
matches must be fixed explicitly.
A changed training profile can require new ICA preparation and decomposition. When the selected range includes the necessary steps, the assigned subject may be rebuilt while unaffected subjects retain compatible reuse. A profile never authorizes stale input reuse or expands the selected range automatically.
Keep branch, override and event-group plan files in your study backup.