bolyai/CIS490
3
0
Fork 0
Code Issues 5 Pull requests Projects 1 Releases Packages Wiki Activity Actions
CIS490/etc
History
Max 2aa7b865fb training/models: knn_semi — semi-supervised self-training KNN 
Registered as `knn_semi`. Answers the research question:

  *If we had ground-truth labels for only a fraction of training
   episodes, could we use the structure of the unlabeled rest to
   recover most of supervised KNN's accuracy?*

Pipeline (Yarowsky-style self-training):

  1. Split train slice deterministically into labeled (label_frac=0.2
     default) and unlabeled (1 - label_frac) by row-index hash.
  2. Fit a "labeler" KNN on the labeled fraction.
  3. Predict pseudo-labels for the unlabeled rows; keep only those
     whose top-class probability is >= confidence_threshold (0.6).
  4. Fit the final KNN on (labeled rows + confident pseudo-labels).
     Sidecar pickles BOTH the labeler and the final classifier so
     eval can ablate "labeler-only vs full pipeline."

Smoke run (567-episode subset, oracle mode, label_frac=0.2):

                       val_macro_f1   test_macro_f1
  knn       (100% labels)   0.737        0.133
  knn_semi  (20% labels)    0.654        0.173

Lower val (less data) but HIGHER cross-device test — pseudo-labeling
acts as a regularizer that prevents overfitting to elliott-thinkpad's
specific neighborhood structure. Honest research finding worth a slide
in the writeup.

Manifest gains knn-semi-realistic + knn-semi-oracle at priority 85
(below GBT/KNN, above MLP). Storage cost = augmented set × n_features
× 4 bytes; same .knn.pkl sidecar format as plain KNN.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-08 13:51:30 -05:00
..
caddy-root.crt bootstrap: auto-issue mTLS leaves to enrolled lab hosts (closes #9, refs #3) 2026-04-30 01:30:29 -05:00
cis490-autoupdate.service lab-host: cis490-autoupdate.timer for self-healing on push 2026-05-01 16:59:31 -05:00
cis490-autoupdate.timer lab-host: cis490-autoupdate.timer for self-healing on push 2026-05-01 16:59:31 -05:00
cis490-bootstrap.service Tier-4 sample source: theZoo (no auth, no operator action) 2026-05-01 01:17:50 -05:00
cis490-cert-fetch.service lab-host: cis490-cert-fetch.timer for automatic mTLS bootstrap retry 2026-05-02 13:30:16 -05:00
cis490-cert-fetch.timer lab-host: cis490-cert-fetch.timer for automatic mTLS bootstrap retry 2026-05-02 13:30:16 -05:00
cis490-dashboard.service training/dashboard: live deck at dashboard.wg, fed by receiver 2026-05-07 21:26:07 -05:00
cis490-doctor-check.service fleet-health: proactive alerts on the Pi + per-host doctor reports 2026-05-02 13:48:31 -05:00
cis490-doctor-check.timer fleet-health: proactive alerts on the Pi + per-host doctor reports 2026-05-02 13:48:31 -05:00
cis490-fleet-health.service fleet-health: proactive alerts on the Pi + per-host doctor reports 2026-05-02 13:48:31 -05:00
cis490-fleet-health.timer fleet-health: proactive alerts on the Pi + per-host doctor reports 2026-05-02 13:48:31 -05:00
cis490-orchestrator.service PIPELINE §5 step 2: canonical manifest at <repo>/manifest.toml 2026-05-04 01:25:01 -05:00
cis490-receiver.service Add receiver: PUT /v1/episodes ingest with sha256 verify and idempotency 2026-04-28 23:34:04 -06:00
cis490-shipper.service shipper: systemd watchdog, quarantine cleanup; doctor surfaces ship errors 2026-05-01 12:02:59 -05:00
cis490-trainer-receiver.service training/fleet: distributed multi-host trainer with capability gating 2026-05-08 01:20:20 -05:00
cis490-trainer-worker.service training/fleet: distributed multi-host trainer with capability gating 2026-05-08 01:20:20 -05:00
lab-host.toml.example etc/lab-host.toml.example: pin Caddy root, not wg-pki client CA (closes #14) 2026-04-30 17:26:36 -05:00
README.md training/dashboard: live deck at dashboard.wg, fed by receiver 2026-05-07 21:26:07 -05:00
receiver.toml.example docs+doctor: surface VERSION-stamp + fallback wiring 2026-05-01 11:54:36 -05:00
training_manifest.toml.example training/models: knn_semi — semi-supervised self-training KNN 2026-05-08 13:51:30 -05:00

README.md

etc/

Templates for system-level files installed by scripts/install-*.sh:

  • cis490-receiver.service — systemd unit for the receiver
  • cis490-dashboard.service — systemd unit for the dashboard.wg live display
  • receiver.toml.example — config template for the receiver
  • cis490-orchestrator.service (TODO) — systemd unit for the orchestrator
  • cis490-shipper.service (TODO) — systemd unit for the shipper
  • lab-host.toml.example (TODO) — config template for the lab host

See docs/deploy.md for the install flow.