code_public/ja4-platform
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(ml): replace logistic regression with MLP fusion and KS drift with ADWIN online learning

Browse Source 
Replace the LogisticRegression meta-learner with a PyTorch MetaFusionMLP
(Linear(3,16)->BN->ReLU->Dropout->Linear(16,1)->Sigmoid) for non-linear
fusion of EIF, NF, and XGBoost scores. Replace KS-test + quantile digest
drift detection with ADWIN (adaptive sliding window, Hoeffding bound).
Replace weekly XGBoost batch retraining with River HoeffdingAdaptiveTree
for incremental online learning (learn_one per cycle). Update all thesis
documentation sections (2.4.2c, 2.4.3, 3.8, discussion, conclusion).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Jacquin Antoine
2026-04-13 16:32:34 +02:00
parent c6cb12981c
commit 7894d39f1c
12 changed files with 502 additions and 306 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/thesis/03_architecture.md
View File

@ -70,7 +70,7 @@
│ │ 4. EIF bifurqué (complet/appli) │ │
│ │ 5. NF log-likelihood scoring │ │
│ │ 6. XGBoost probabilité │ │
│ │ 7. Fusion LR fusion │ │
│ │ 7. Meta-Model MLP fusion │ │
│ │ 8. HDBSCAN clustering (NF latent) │ │
│ │ 9. Écriture résultats ClickHouse │ │
│ └──────────────────────────────────┘ │
@ -240,7 +240,7 @@ Session entrante
├── asn_label == 'human' ?
│ ── OUI → baseline EIF training (sans étiquette bot)
│
└── Sinon → Triple-voix : EIF + NF + XGBoost + Fusion LR
└── Sinon → Triple-voix : EIF + NF + XGBoost + Meta-Model Stacking (MLP non-linéaire)
```
#### Seuil adaptatif