In high-dimensional regression problems, often a relatively small subset of the features are relevant for predicting the outcome, and methods that impose sparsity on the solution are popular. When multiple correlated ...
详细信息
In high-dimensional regression problems, often a relatively small subset of the features are relevant for predicting the outcome, and methods that impose sparsity on the solution are popular. When multiple correlated outcomes are available (multitask), reduced rank regression is an effective way to borrow strength and capture latent structures that underlie the data. Our proposal is motivated by the UK Biobank population-based cohort study, where we are faced with large-scale, ultrahigh-dimensional features, and have access to a large number of outcomes (phenotypes)-lifestyle measures, biomarkers, and disease outcomes. We are hence led to fit sparse reduced-rank regression models, using computational strategies that allow us to scale to problems of this size. We use a scheme that alternates between solving the sparse regression problem and solving the reduced rank decomposition. For the sparse regression component we propose a scalable iterative algorithm based on adaptive screening that leverages the sparsity assumption and enables us to focus on solving much smaller subproblems. The full solution is reconstructed and tested via an optimality condition to make sure it is a valid solution for the original problem. We further extend the method to cope with practical issues, such as the inclusion of confounding variables and imputation of missing values among the phenotypes. Experiments on both synthetic data and the UK Biobank data demonstrate the effectiveness of the method and the algorithm. We present multiSnpnet package, available at http://***/junyangq/multiSnpnet that works on top of PLINK2 files, which we anticipate to be a valuable tool for generating polygenic risk scores from human genetic studies.
In the shape optimization analysis of electrostatically controlled deployable membrane reflector (ECDMR), the poor balance of the membrane with spatial skirt cable is the main reason for the degradation of the reflect...
详细信息
In the shape optimization analysis of electrostatically controlled deployable membrane reflector (ECDMR), the poor balance of the membrane with spatial skirt cable is the main reason for the degradation of the reflector surface precision. At the same time, while in the optimization of the pre-stress of large deployable reflectors, the membrane structure has strong geometric nonlinearity, and the lack of necessary gradient information leads to low efficiency of the calculation. Therefore, according to the structural characteristics of the cable-membrane reflector, the two optimization objectives of the highest surface precision and the most uniform stress distribution is considered, and a hierarchical shape optimization method is proposed based on membrane theory and unbalance force analysis, which realizes the solution of the multi-objective optimization problem when the three optimization objectives have different priorities. The interior effective membrane reflective surface is analyzed using membrane theory, and the external membrane with spatial skirt cable is analyzed using nonlinear finite element method. The gradient information required for optimization is also obtained by unbalance force analysis, which ensures better optimization efficiency and can solve large-scale optimization problems. Therefore, the node positions and element equilibrium stresses of the structure can be optimized simultaneously, which further expands the search space of the equilibrium state and reduces the possibility of falling into local optimal solutions. Finally, a numerical example of a 30 m aperture ECDMR is calculated and analyzed, and the results show that the proposed method can obtain a high surface precision with a fast solution speed.
In this work, we present a modification (refinement) of the ensemble-based method for constrained waterflooding optimization. The problem of determining life-cycle rate controls for both producer and injector wells th...
详细信息
In this work, we present a modification (refinement) of the ensemble-based method for constrained waterflooding optimization. The problem of determining life-cycle rate controls for both producer and injector wells that maximize the net present value, NPV, subject to well and field-wide capacity constraints is formulated and solved using sequential quadratic programming, SQP. The required gradient is approximately computed by an ensemble-based method. Field NPV is decomposed as the sum of the NPVs of each well. Sensitivity matrix of well NPVs with respect to controls of all wells is obtained from ensemble-based covariance matrices of controls and of well NPVs to controls. For efficiency reasons, ensemble size should be kept small which results in sampling errors. The approximate gradient is the sum of the columns of the refined sensitivity matrix. Using small-sized ensembles introduces spurious correlations that degrade gradient quality. Novel nondistance-based localization technique is employed to mitigate the deleterious effects of spurious correlations to refine the sensitivity of NPV of production wells with respect to injector controls. The localization technique is based on the connectivity of each injector/producer pair using a producer-based capacitance resistance model (CRMP). Competitiveness coefficients are developed to refine sensitivity of NPV of production wells with respect to producer controls, obtained using an interference test. A new procedure is proposed for consideration of maximum water-cut limit resulting in producer shut-in during the optimization process. Smoothing techniques are also introduced to avoid excessive abrupt jumps in well controls and to improve the overall optimization efficiency. Procedures and refinements are applied to a realistic reservoir taken from the literature, TNO Brugge Field, to demonstrate the resulting level of objective function improvement and variability reduction of the obtained solutions. NPV solution statist
暂无评论