WiSleep: Smartphone-driven Sleep Population Monitoring with Unsupervised Learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
With sleep deprivation being a public health concern, sleep monitoring technology, mainly through consumer-grade wearables, has shown value among users to better understand their most fundamental measure of health. Unfortunately, utilizing wearable technology is bound to the conditions of users owning these devices and using them at bedtime every night. While wearables can deliver highly personalized sleep insights to users, they inadvertently affect the ability of sleep monitoring solutions to reach unprivileged sections of society due to added costs and device accessibility. With our primary motivation to promote sleep monitoring for public health use cases at the population scale, we developed WiSleep , a sleep monitoring system that infers sleep duration from solely relying on a user’s smartphone without requiring a wearable device. Unlike prior efforts that use supervised learning methods and require labeled training data to train sleep models, our method is based on unsupervised learning, which enables easy deployment to new population groups or new regions without a need for labeled data collection and training. Specifically, we employ the smartphone activity of the user, represented by time series of WiFi network event rates, as input data to infer the user’s sleep duration (i.e., sleep time and wake time) through an unsupervised Bayesian change point detection ensemble model. Our evaluation shows WiSleep ’s efficacy in being a low-cost accessible sleep monitoring approach. We present results that yield comparable performance to prior techniques, particularly those requiring new users’ labeled data to achieve model personalization. System evaluation from a user study achieved an average of 93.68% accuracy within 59 minutes of sleep time error, 31 minutes of wake time error, and 57 minutes of sleep duration error by utilizing coarse-grained time series data. We demonstrate the application of our technique to predict sleep for 1,000 anonymous users and enable population-scale analytics with low computational overhead.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.004 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it