We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Medical data wrangling with sequential variational autoencoders

Abstract: Medical data sets are usually corrupted by noise and missing data. These missing patterns are commonly assumed to be completely random, but in medical scenarios, the reality is that these patterns occur in bursts due to sensors that are off for some time or data collected in a misaligned uneven fashion, among other causes. This paper proposes to model medical data records with heterogeneous data types and bursty missing data using sequential variational autoencoders (VAEs). In particular, we propose a new methodology, the Shi-VAE, which extends the capabilities of VAEs to sequential streams of data with missing observations. We compare our model against state-of-the-art solutions in an intensive care unit database (ICU) and a dataset of passive human monitoring. Furthermore, we find that standard error metrics such as RMSE are not conclusive enough to assess temporal models and include in our analysis the cross-correlation between the ground truth and the imputed signal. We show that Shi-VAE achieves the best performance in terms of using both metrics, with lower computational complexity than the GP-VAE model, which is the state-of-the-art method for medical records.
Comments: Accepted for publication in IEEE Journal of Biomedical and Health Informatics (JBHI)
Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML)
DOI: 10.1109/JBHI.2021.3123839
Cite as: arXiv:2103.07206 [cs.LG]
  (or arXiv:2103.07206v2 [cs.LG] for this version)

Submission history

From: Daniel Barrejon [view email]
[v1] Fri, 12 Mar 2021 10:59:26 GMT (376kb,D)
[v2] Mon, 8 Nov 2021 09:21:34 GMT (837kb,D)

Link back to: arXiv, form interface, contact.