# Separating Expert Retention from Autonomous Source Inference in Raw-ECG-Replay-Free Continual ECG Deployment

arXiv: https://arxiv.org/abs/2607.01674 · HTML: https://mlsanity.com/paper/2607.01674/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 23 references that are in the map.
- First posted on arXiv 2 Jul 2026.

## Builds on (most cited first, 12 of 12)

- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [Learning without Forgetting](https://mlsanity.com/paper/1606.09282/) (2016)
- [iCaRL: Incremental Classifier and Representation Learning](https://mlsanity.com/paper/1611.07725/) (2016)
- [Learning to Prompt for Continual Learning](https://mlsanity.com/paper/2112.08654/) (2021)
- [Long-tail learning via logit adjustment](https://mlsanity.com/paper/2007.07314/) (2020)
- [CODA-Prompt: COntinual Decomposed Attention-based Prompting for Rehearsal-Free Continual Learning](https://mlsanity.com/paper/2211.13218/) (2022)
- [Balanced Meta-Softmax for Long-Tailed Visual Recognition](https://mlsanity.com/paper/2007.10740/) (2020)
- [Expert Gate: Lifelong Learning with a Network of Experts](https://mlsanity.com/paper/1611.06194/) (2016)
- [Continual Learning Based on OOD Detection and Task Masking](https://mlsanity.com/paper/2203.09450/) (2022)
- [Dynamic Prototype Rehearsal for Continual ECG Arrhythmia Detection](https://mlsanity.com/paper/2501.07555/) (2025)
- [ECG-CL: A Comprehensive Electrocardiogram Interpretation Method Based on Continual Learning](https://mlsanity.com/paper/2304.04646/) (2023)
- [MELEP: A Novel Predictive Measure of Transferability in Multi-Label ECG Diagnosis](https://mlsanity.com/paper/2311.04224/) (2023)

## Abstract (arXiv)

In multi-source ECG deployment, new sources may arrive when earlier raw ECGs cannot be retained or replayed. Isolating source-specific classifiers on a frozen backbone prevents parameter interference, but source-unknown inference still requires selecting an appropriate expert. We study this distinction with IRFE-ECG, a controlled continual-deployment framework built on frozen 1024-dimensional ECGFounder features. Each arriving source adds an isolated Balanced-Softmax linear expert, while a lightweight router is re-fitted using retained frozen training features and source labels from previously observed sources. Rather than proposing a new routing architecture, the main contribution is to separate preserved expert performance from autonomous source inference and quantify the resulting deployment gap. Across CPSC, PTB-XL, Georgia, and Chapman-Shaoxing, source-aware expert selection reaches 0.7915 ± 0.0036 Macro-F1, close to a matched offline independent-head reference at 0.7885 ± 0.0009. Without source IDs, an MLP router reaches 0.7756 ± 0.0027, while top-2 margin fusion reaches 0.7782 ± 0.0022. The top-2 improvement is small (+0.0026) and not statistically significant under paired bootstrap. Across three domain orders, the top-2-to-oracle gap remains 0.0111-0.0133, indicating a persistent source-inference gap within this protocol. The results are record-level because reliable patient identifiers were unavailable. The method replays no raw ECGs, but it retains frozen feature vectors for router updates and is therefore raw-ECG-replay-free rather than memory-free. Code is publicly available at https://github.com/yufanlu221/IRFE-ECG.