Search

Large training datasets are important for deep learning-based methods. For medical image segmentation, it could be however difficult to obtain large number of labeled training images solely from one center. Distributed learning, such as swarm learning, has the potential to use multi-center data without breaching data privacy. However, data distributions across centers can vary a lot due to the diverse imaging protocols and vendors (known as feature skew). Also, the regions of interest to be segmented could be different, leading to inhomogeneous label distributions (referred to as label skew). With such non-independently and identically distributed (Non-IID) data, the distributed learning could result in degraded models. In this work, we propose a novel swarm learning approach, which assembles local knowledge from each center while at the same time overcomes forgetting of global knowledge during local training. Specifically, the approach first leverages a label skew-awared loss to preserve the global label knowledge, and then aligns local feature distributions to consolidate global knowledge against local feature skew. We validated our method in three Non-IID scenarios using four public datasets, including the Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation (M&Ms) dataset, the Federated Tumor Segmentation (FeTS) dataset, the Multi-Modality Whole Heart Segmentation (MMWHS) dataset and the Multi-Site Prostate T2-weighted MRI segmentation (MSProsMRI) dataset. Results show that our method could achieve superior performance over existing methods. Code will be released via https://zmiclab.github.io/projects.html once the paper gets accepted.

Objective: To develop a New Chinese Diabetes Risk Score for screening undiagnosed type 2 diabetes in China., Research Design and Methods: Data from the China National Diabetes and Metabolic Disorders Study conducted from June 2007 to May 2008 comprising 16,525 men and 25,284 women aged 20-74 years were analyzed. Undiagnosed type 2 diabetes was detected based on fasting plasma glucose ≥7.0 mmol/L or 2-h plasma glucose ≥11.1 mmol/L in people without a prior history of diabetes. β-Coefficients derived from a multiple logistic regression model predicting the presence of undiagnosed type 2 diabetes were used to calculate the New Chinese Diabetes Risk Score. The performance of the New Chinese Diabetes Risk Score was externally validated in two studies in Qingdao: one is prospective with follow-up from 2006 to 2009 (validation 1) and another cross-sectional conducted in 2009 (validation 2)., Results: The New Chinese Diabetes Risk Score includes age, sex, waist circumference, BMI, systolic blood pressure, and family history of diabetes. The score ranges from 0 to 51. The area under the receiver operating curve of the score for undiagnosed type 2 diabetes was 0.748 (0.739-0.756) in the exploratory population, 0.725 (0.683-0.767) in validation 1, and 0.702 (0.680-0.724) in validation 2. At the optimal cutoff value of 25, the sensitivity and specificity of the score for predicting undiagnosed type 2 diabetes were 92.3 and 35.5%, respectively, in validation 1 and 86.8 and 38.8% in validation 2., Conclusions: The New Chinese Diabetes Risk Score based on nonlaboratory data appears to be a reliable screening tool to detect undiagnosed type 2 diabetes in Chinese population.

This paper presents an efficient algorithmic implementation of the iterative minimization formulation (IMF) for fast local search of transition state on potential energy surface. The IMF is a second order iterative scheme providing a general and rigorous description for the eigenvector-following (min-mode following) methodology. We offer a unified interpretation in numerics via the IMF for existing eigenvector-following methods, such as the gentlest ascent dynamics, the dimer method and many other variants. We then propose our new algorithm based on the IMF. The main feature of our algorithm is that the translation step is replaced by solving an optimization subproblem associated with an auxiliary objective function which is constructed from the min-mode information. We show that using an efficient scheme for the inexact solver and enforcing an adaptive stopping criterion for this subproblem, the overall computational cost will be effectively reduced and a super-linear rate between the accuracy and the computational cost can be achieved. A series of numerical tests demonstrate the significant improvement in the computational efficiency for the new algorithm. [ABSTRACT FROM AUTHOR]

Biclustering (co-clustering, two-mode clustering), as one of the classical unsupervised learning methods, has been applied in many different fields in recent years. Different types of biclustering methods have been developed such as probabilistic methods, two-way clustering methods, variance minimization methods, and so on. However, few regression-based methods have been proposed to the best of our knowledge. Such methods have been applied in traditional clustering, which can improve both the computational efficiency and the clustering accuracy. In this paper, we present a penalized regression-based method for localizing the biclusters (PRbiclust). By imposing Truncated LASSO Penalty (TLP) and group TLP terms to penalize the column vectors and the row vectors in the regression model, the structure of biclusters in the data matrix is recovered. The model is formulated as an optimization problem with nonconvex penalties, and a computationally efficient algorithm is proposed to solve it. Convergence of the algorithm is proved. To extract the biclusters from the recovered data matrix, we propose a graph-based localization method. An evaluation criterion is also proposed to measure the efficiency of bicluster localization when noise entries exist. We apply the proposed method to both simulated datasets with different setups and a real dataset. Experiments show that this method can well capture the bicluster structure, and performs better than the existing works. [ABSTRACT FROM AUTHOR]

Abstract: Plane wave pseudopotential (PWP) density functional theory (DFT) calculation is the most widely used method for material simulations, but its absolute speed stagnated due to the inability to use large scale CPU based computers. By a drastic redesign of the algorithm, and moving all the major computation parts into GPU, we have reached a speed of 12s per molecular dynamics (MD) step for a 512 atom system using 256GPU cards. This is about 20 times faster than the CPU version of the code regardless of the number of CPU cores used. Our tests and analysis on different GPU platforms and configurations shed lights on the optimal GPU deployments for PWP-DFT calculations. An 1800 step MD simulation is used to study the liquid phase properties of GaInP. [Copyright &y& Elsevier]