ulrica66666/TS-MEMTS: v1.0.0 — MEMTS implementation release for KDD 2026 camera-ready

Initial code release accompanying the KDD 2026 paper:

MEMTS: Internalizing Domain Knowledge via Parameterized Memory for Retrieval-Free Domain Adaptation of Time Series Foundation Models (DOI: 10.1145/3770855.3818032).

This release contains the implementation of:

• Knowledge Persistence Module (KPM) — Context Encoder + Future-Separable Decoder.
• Adaptive Fusion (AF) — Gated residual correction mechanism.
• Retrieval-Free Inference — Parametric forward pass replacing online database queries.
• Knowledge base construction — Faiss-based retrieval logic for offline knowledge internalization.
• End-to-end pipeline runner — run.sh.

Pre-trained KPM and AF checkpoints, the ECL dataset slice, and Faiss knowledge base indices will be released separately. See README.md for details.