Density-sensitive semisupervised inference

Azizyan, Martin; Singh, Aarti; Wasserman, Larry

doi:10.1214/13-aos1092

Cited by 18 publications

(20 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our setting, determining whether or not a complication occurred for a given case often requires extensive manual review of the case file and is subject to varying clinical definitions of outcomes. A promising modeling approach for this situation is semisupervised inference, such as discussed in Azizyan et al 22 Existing formal methods for semisupervised inference typically make assumptions linking the probability density function of the predictor variables to the function used for outcome prediction. It is unclear if such methods would be effective in our setting with a high-dimensional predictor set and low fraction of labeled data; we have not pursued this approach here.…”

Section: Complication Outcome Variablesmentioning

confidence: 99%

Leveraging electronic health records for predictive modeling of post-surgical complications

Weller¹,

Lovely

Larson

et al. 2017

Stat Methods Med Res

View full text Add to dashboard Cite

Hospital-specific electronic health record systems are used to inform clinical practice about best practices and quality improvements. Many surgical centers have developed deterministic clinical decision rules to discover adverse events (e.g. postoperative complications) using electronic health record data. However, these data provide opportunities to use probabilistic methods for early prediction of adverse health events, which may be more informative than deterministic algorithms. Electronic health record data from a set of 9598 colorectal surgery cases from 2010 to 2014 were used to predict the occurrence of selected complications including surgical site infection, ileus, and bleeding. Consistent with previous studies, we find a high rate of missing values for both covariates and complication information (4-90%). Several machine learning classification methods are trained on an 80% random sample of cases and tested on a remaining holdout set. Predictive performance varies by complication, although an area under the receiver operating characteristic curve as high as 0.86 on testing data was achieved for bleeding complications, and accuracy for all complications compares favorably to existing clinical decision rules. Our results confirm that electronic health records provide opportunities for improved risk prediction of surgical complications; however, consideration of data quality and consistency standards is an important step in predictive modeling with such data.

show abstract

Section: Complication Outcome Variablesmentioning

confidence: 99%

Leveraging electronic health records for predictive modeling of post-surgical complications

Weller¹,

Lovely

Larson

et al. 2017

Stat Methods Med Res

View full text Add to dashboard Cite

show abstract

“…It is known that the affinity between data points depends not only on the location but also on the neighborhoods of two points. The density of data has been considered by various metrics in supervised [7], unsupervised [8], semisupervised [9] and deep learning [10]. The affinity between two data points with different neighborhood density is different from the affinity when both neighborhoods have the same density [11].…”

Section: Related Workmentioning

confidence: 99%

Cross-covariance based affinity for graphs

et al. 2020

View full text Add to dashboard Cite

The accuracy of graph based learning techniques relies on the underlying topological structure and affinity between data points, which are assumed to lie on a smooth Riemannian manifold. However, the assumption of local linearity in a neighborhood does not always hold true. Hence, the Euclidean distance based affinity that determines the graph edges may fail to represent the true connectivity strength between data points. Moreover, the affinity between data points is influenced by the distribution of the data around them and must be considered in the affinity measure. In this paper, we propose two techniques, CCGA L and CCGA N that use cross-covariance based graph affinity (CCGA) to represent the relation between data points in a local region. CCGA L also explores the additional connectivity between data points which share a common local neighborhood. CCGA N considers the influence of respective neighborhoods of the two immediately connected data points, which further enhance the affinity measure. Experimental results of manifold learning on synthetic datasets show that CCGA is able to represent the affinity measure between data points more accurately. This results in better low dimensional representation. Manifold regularization experiments on standard image dataset further indicate that the proposed CCGA based affinity is able to accurately identify and include the influence of the data points and its common neighborhood that increase the classification accuracy. The proposed method outperforms the existing state-of-the-art manifold regularization methods by a significant margin.

show abstract

“…A density of samples in a data space has been already taken into account by various metrics used for different machine learning tasks: unsupervised learning (Soleimani et al 2015); supervised learning (Plant et al 2006or Aggarwal 2007; semi-supervised learning (Azizyan et al 2013); reinforcement learning (Rojanaarpa and Kataeva 2016); deep learning (Nicolau and McDermott 2016); matchmaking for queries and ontologies (Naumenko et al 2006). The common assumption is that the two data samples x and y from the regions with different density of neighboring data samples must be considered more distant than in the case when both regions have the same density; and therefore some penalty is applied to a spatial distance between them.…”

Section: Introductionmentioning

confidence: 99%

Social Distance metric: from coordinates to neighborhoods

Terziyan

2017

International Journal of Geographical Information Science

View full text Add to dashboard Cite

Choice of a distance metric is a key for the success in many machine learning and data processing tasks. The distance between two data samples traditionally depends on the values of their attributes (coordinates) in a data space. Some metrics also take into account the distribution of samples within the space (e.g. local densities) aiming to improve potential classification or clustering performance. In this paper, we suggest the Social Distance metric that can be used on top of any traditional metric. For a pair of samples x and y, it averages the two numbers: the place (rank), which sample y holds in the list of ordered nearest neighbors of x; and vice versa, the rank of x in the list of the nearest neighbors of y. Average is a contraharmonic Lehmer mean, which penalizes the difference between the numbers by giving values greater than the Arithmetic mean for the unequal arguments. We consider normalized average as a distance function and we prove it to be a metric. We present several modifications of such metric and show that their properties are useful for a variety of classification and clustering tasks in data spaces or graphs in a Geographic Information Systems context and beyond.

show abstract

Density-sensitive semisupervised inference

Cited by 18 publications

References 18 publications

Leveraging electronic health records for predictive modeling of post-surgical complications

Leveraging electronic health records for predictive modeling of post-surgical complications

Cross-covariance based affinity for graphs

Social Distance metric: from coordinates to neighborhoods

Contact Info

Product

Resources

About