cd /news/machine-learning/tibecg-temporal-information-bottlene… · home › topics › machine-learning › article
[ARTICLE · art-171570] src=arxiv.org ↗ pub= topic=machine-learning verified=true sentiment=· neutral

TIBECG: Temporal Information Bottleneck for Lead-Specific Demographic Privacy in ECG Waveform Release

A 2,302,337-parameter temporal information bottleneck processor, TIBECG, retained at least 95% of each lead's clean macro-F1 across 6,386 ECG recordings, ten folds, 12 leads and 21 latent lengths while suppressing age and sex inference, according to an arXiv paper (2611.21694v1). TIBECG's weaker age/sex interval gain exceeded the best eligible conventional transformation on all leads by a median 22.08 percentage points, and the authors report that direct downsampling usually suppresses demographic inference more strongly but at a larger loss of diagnostic utility. The authors state the findings support lead-specific evaluation of waveform processing, not validated consumer deployment or expert-readable reconstruction, and note that lower latent Kullback-Leibler divergence does not reliably rank residual sex leakage at a fixed length.

key takeaway

TIBECG's temporal information bottleneck kept at least 95% of clean macro-F1 per ECG lead while beating the best conventional transformation on age/sex suppression by a median 22.08 percentage points, but lower latent KL divergence did not reliably rank residual sex leakage at a fixed length.

read1 min views2 publishedNov 18, 2026

arXiv:2611.21694v1 Announce Type: new Abstract: Transmitting an electrocardiogram (ECG) waveform can expose demographic information beyond an intended diagnostic task. We evaluate TIBECG, an architectural temporal information bottleneck, on individual leads extracted from hospital twelve-lead recordings. A 2,302,337-parameter processor varies its latent length while retaining the same nine-score consistency objective, without age or sex labels in protector fitting. Across 6,386 recordings, ten folds, 12 leads and 21 lengths, retrospective selections retained at least 95% of each lead's clean macro-F1. Their weaker age/sex interval gain exceeded the best eligible conventional transformation on all leads, by a median 22.08 percentage points. The implemented sampling-rate comparison shows that direct downsampling usually suppresses demographic inference more strongly but at a larger loss of diagnostic utility, whereas TIBECG favors utility preservation. Selection stability also varies by lead. Lower latent Kullback-Leibler (KL) divergence does not reliably rank residual sex leakage at a fixed length. These findings support lead-specific evaluation of waveform processing, not validated consumer deployment or expert-readable reconstruction. When the waveform-input diagnostic classifier is hosted as part of a proprietary cloud service rather than deployed on the device, waveform transmission is required to obtain its scores.

── more in #machine-learning 4 stories · sorted by recency
── more on @tibecg 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/tibecg-temporal-info…] indexed:0 read:1min 2026-11-18 · —