RC-KT reproducibility workflow

[skim0119]

Summary

Collect the publication-grade RC-KT reproducibility workflow: state embedding, reusable MiV operators, full-cohort analysis scripts, launchers, tests, and documentation.

This integration PR is publication-gated and will merge into main upon publication. The main README contains the statistical-diagnostics graph; this branch adds the RC and KT-RC graphs and links the complete workflow under examples/rc_kt/.

Subordinate work

• External hierarchical GPFA prerequisite: https://github.com/skim0119/MiV-State-Embedding/pull/3
• Optional direct dependency: Add optional RC-KT analysis dependencies #588
• Reusable RC-KT operators and numerical corrections: Add RC-KT analysis operators #590
• Reproducibility workflow: Add full RC-KT reproducibility workflow #591

The miv-state-space dependency temporarily tracks the feat/rc-kt-hgpfa branch; package publication will be handled separately.

Knowledge-transfer graph

Expert and student recordings are responses to the same experimental input protocol. The public MiV graph pairs their trial streams once and keeps selection, GPFA, latent projection, readout, and transplant as explicit nodes:

expert_trials >> kt_input
student_trials >> kt_input
kt_input >> expert_trial_stream >> expert_gpfa
kt_input >> student_trial_stream >> student_gpfa
expert_gpfa >> student_gpfa
expert_gpfa >> expert_features >> expert_readout
student_gpfa >> student_features
expert_features >> transplant
student_features >> transplant
expert_readout >> transplant

Pipeline(transplant).run(...)
kt_result = transplant.output()

The only required serial dependency is expert_gpfa >> student_gpfa, which passes learned expert kernel parameters into the student observation fit. Each node otherwise has its own input/output boundary, cache identity, and applicable callback. KnowledgeTransfer only aligns already projected latent states and an already fitted readout, then returns the immediate transplant plus optional prior-centered refinement.

Review map

• Review Add RC-KT analysis operators #590 for numerical corrections, operator APIs, result dataclasses, cache identities, and paired-input KT orchestration.
• Review Add full RC-KT reproducibility workflow #591 for manifest handling, full-cohort scripts, publication outputs, Parsl distribution, and Slurm launchers.
• Real data and cluster execution are intentionally deferred because the dataset and private cluster configuration are not bundled.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.25%. Comparing base (3eb199b) to head (7030e3b).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #589   +/-   ##
=======================================
  Coverage   86.25%   86.25%           
=======================================
  Files          28       28           
  Lines        1157     1157           
  Branches      115      115           
=======================================
  Hits          998      998           
  Misses        134      134           
  Partials       25       25           

Flag	Coverage Δ
mpi-tests	86.25% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.