SUSTain: Scalable Unsupervised Scoring for Tensors and its Application to Phenotyping

By Ioakeim Perros, Evangelos E. Papalexakis, Haesun Park, R. Vuduc, Xiaowei Yan, Christopher deFilippi, Walter F. Stewart, Jimeng Sun. Knowledge Discovery and Data Mining, 2018

589 citations recorded by the source corpus.

This paper presents a new method, which we call SUSTain, that extends real-valued matrix and tensor factorizations to data where values are integers. Such data are common when the values correspond to event counts or ordinal measures. The conventional approach is to treat integer data as real, and then apply real-valued factorizations. However, doing so fails to preserve important characteristics of the original data, thereby making it hard to interpret the results. Instead, our approach extracts factor values from integer datasets as scores that are constrained to take values from a small integer set. These scores are easy to interpret: a score of zero indicates no feature contribution and higher scores indicate distinct levels of feature importance. At its core, SUSTain relies on: a) a problem partitioning into integer-constrained subproblems, so that they can be optimally solved in an efficient manner; and b) organizing the order of the subproblems' solution, to promote reuse of shared intermediate results. We propose two variants, SUSTain_M and SUSTain_T, to handle both matrix and tensor inputs, respectively. We evaluate SUSTain against several state-of-the-art baselines on both synthetic and real Electronic Health Record (EHR) datasets. Comparing to those baselines, SUSTain shows either significantly better fit or orders of magnitude speedups that achieve a comparable fit (up to 425× faster). We apply SUSTain to EHR datasets to extract patient phenotypes (i.e., clinically meaningful patient clusters). Furthermore, 87% of them were validated as clinically meaningful phenotypes related to heart failure by a cardiologist.

Subfields: Artificial Intelligence, Computational Mathematics, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/370efc8d4251fb625d80b5e560965cf11282805f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xin Li
2. Yarin Gal

Citations in the corpus

• Current status and future perspectives of lithium metal batteries (2020, background): “for example, Lithium-Sulfur (Li-S) and Li-O2 batteries, has substantially advanced in the past decade [20,21].”
• COPA: Constrained PARAFAC2 for Sparse & Large Datasets (2018, unknown): “Tensor factorization encompasses a set of powerful analytic methods that have been successfully applied in many application domains: social network analysis [1, 2], urban planning [3], and health analytics [4, 5, 6, 7, 8].”
• Constructing low-cost Ni3C/twin-crystal Zn0.5Cd0.5S heterojunction/homojunction nanohybrids for efficient photocatalytic H2 evolution (2021, background): “As generally perceived, nanostructured sulfide-based photocatalysts, such as zero-dimensional nanoparticle (NP)/ nanocrystals, one-dimensional (1D) nanowire/nanorods, two-dimensional nanoplates/nanosheets [9], and three-dimensional nanoflowers/nanospheres [10,11] can significantly shorten the migrat”
• Microbial fuel cells: From fundamentals to applications. A review (2017, background): “pathways (acidic and alkaline) that were previously well described in terms of reactions and relative redox potentials [219,319,320].”
• Monolithic Perovskite Tandem Solar Cells: A Review of the Present Status and Advanced Characterization Methods Toward 30% Efficiency (2020, background): “[322] For sub-cells with rather high and equal FFs of the sub-cells, the tandem FF can be even higher than the individual sub-cell’s FFs and is a function of mismatch[68] (see also the lower panel of Figure 12).”
• Recent progress of biomass-derived carbon materials for supercapacitors (2020, background): “hemicellulose-derived carbon electrode materials for supercapacitor applications [21].”
• Ionic thermoelectric materials and devices (2021, background): “The majority of the world’s power is generated by heat engines with efficiency of only 30%–40% [1,2], which means that at least 60% of the consumed fossil fuel dissipates typically in the form of waste heat.”
• A review of lithium-ion battery safety concerns: The issues, strategies, and testing standards (2020, background): “Furthermore, the poor thermal stability of these electrolytes causes their decomposition and reactions with the electrode’s active materials [117].”