This paper presents a new approach to model selection based on hypothesis testing. We ÿrst describe a procedure to generate di erent scores for any candidate model from a single sample of training data and then discuss how to apply multiple comparison procedures (MCP) to model selection. MCP statistical tests allow us to compare three or more groups of data while controlling the probability of making at least one Type I error. The complete procedure is illustrated on several model selection tasks, including the determination of the number of hidden units for feed-forward neural networks and the number of kernels for RBF networks.
High angle annular dark field scanning transmission electron microscopy (HAADF-STEM) is a powerful tool to quantify size, shape, position, and composition of nano-objects with the assessment of image simulation. Due to the high computational requirements needed, nowadays it can only be applied to a few unit cells in standard computers. To overpass this limitation, a parallel software (SICSTEM) has been developed. This software can afford HAADF-STEM image simulations of nanostructures composed of several hundred thousand atoms in manageable time. The usefulness of this tool is exemplified by simulating a HAADF-STEM image of an InAs nanowire.
The compositional distribution in a self-assembled InAs(P) quantum wire grown by molecular beam epitaxy on an InP(001) substrate has been determined by electron energy loss spectrum imaging. We have determined the strain and stress fields generated in and around this wire capped with a 5 nm InP layer by finite element calculations using as input the compositional map experimentally obtained. Preferential sites for nucleation of wires grown on the surface of this InP capping layer are predicted, based on chemical potential minimization, from the determined strain and stress fields on this surface. The determined preferential sites for wire nucleation agree with their experimentally measured locations. The method used in this paper, which combines electron energy loss spectroscopy, high-resolution Z contrast imaging, and elastic theory finite element calculations, is believed to be a valuable technique of wide applicability for predicting the preferential nucleation sites of epitaxial self-assembled nano-objects.
In this paper a technique to design three dimensional (3D) devices to focus acoustic waves composed of scattering elements is proposed. The devices are designed and optimized in two dimensions (2D) with the help of a genetic algorithm and the 2D multiple scattering formalism. The transition from 2D to 3D is made by applying a rotation operation to the optimized design, thus passing from a set of 2D circular scatters to their equivalent 3D concentric rings of circular section and finite dimensions, considerably improving its performance. The method has been applied to the design and theoretical characterization of a single-focus acoustic lens and a tunable lens capable of changing the focal length with frequency. A prototype lens was fabricated using aluminum rings clamped to a rigid frame, obtaining a good agreement between theory and experiment.
To train, evaluate, and validate the application of a deep learning framework in three-dimensional ultrasound (3D US) for the automatic segmentation of ventricular volume in preterm infants with post haemorrhagic ventricular dilatation (PHVD). We trained a 2D convolutional neural network (CNN) for automatic segmentation ventricular volume from 3D US of preterm infants with PHVD. The method was validated with the Dice similarity coefficient (DSC) and the intra-class coefficient (ICC) compared to manual segmentation. The mean birth weight of the included patients was 1233.1 g (SD 309.4) and mean gestational age was 28.1 weeks (SD 1.6). A total of 152 serial 3D US from 10 preterm infants with PHVD were analysed. 230 ventricles were manually segmented. Of these, 108 were used for training a 2D CNN and 122 for validating the methodology for automatic segmentation. The global agreement for manual versus automated measures in the validation data (n = 122) was excellent with an ICC of 0.944 (0.874–0.971). The Dice similarity coefficient was 0.8 (± 0.01). 3D US based ventricular volume estimation through an automatic segmentation software developed through deep learning improves the accuracy and reduces the processing time needed for manual segmentation using VOCAL. 3D US should be considered a promising tool to help deepen our current understanding of the complex evolution of PHVD.
We show in this article that it is possible to obtain elemental compositional maps and profiles with atomic-column resolution across an InxGa1−xAs multilayer structure from 5th-order aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images. The compositional profiles obtained from the analysis of HAADF-STEM images describe accurately the distribution of In in the studied multilayer in good agreement with Muraki's segregation model [Muraki, K., Fukatsu, S., Shiraki, Y. & Ito, R. (1992). Surface segregation of In atoms during molecular beam epitaxy and its influence on the energy levels in InGaAs/GaAs quantums wells. Appl Phys Lett61, 557–559].
Atomic steps at growth surfaces are important heterogeneous sources for nucleation of epitaxial nano-objects. In the presence of misfit strain, we show that the nucleation process takes place preferentially at the upper terrace of the step as a result of the local stress relaxation. Evidence for strain-induced nucleation comes from the direct observation by postgrowth, atomic resolution, Z-contrast imaging of an InAs-rich region in a nanowire located on the upper terrace surface of an interfacial diatomic step. © 2007 American Institute of Physics. ͓DOI: 10.1063/1.2790483͔Atomic steps located at the surfaces of substrates play a central role in controlling the growth mechanism of a wide range of materials. The effect of such steps on the growth process, the morphology, the stress and strain distributions, and the functionality of the materials has been extensively investigated. [1][2][3][4][5][6][7][8][9][10][11] In particular, these steps are thought to constitute a heterogeneous source of nucleation for the formation of nano-objects and structural defects at the initial stages of the growth of many semiconductor nanostructures. 12,13 It is critically important to identify the nucleation sources for individual nano-objects such as quantum dots and wires because they constitute the fundamental blocks of future nanoelectronics and nanophotonics devices. 14 The ability to control the nucleation of these nano-objects will contribute significantly to the development of reliable single-photon sources for applications in quantum information technology. [15][16][17] Nanowires constitute a special case of self-assembled nano-objects. 18 Self-assembled nano-objects can be formed spontaneously via the Stranski-Krastanov growth mode for certain semiconductor materials when a few monolayers are epitaxially deposited on a lattice-mismatched substrate. 19,20 Strain is widely accepted to be the driving force for the nucleation of these nano-objects, but as yet there has been no direct experimental evidence for the role of steps, and their associated stress enhancement, in the self-organized growth of nanowire arrays. 21,22 In this letter we show the direct imaging, by aberrationcorrected Z-contrast scanning transmission electron microscopy ͑STEM͒ and atomic force microscopy, that, in the presence of misfit strain, nanowires preferentially nucleate on the upper terrace of a diatomic step. With finite element elasticity calculations, we demonstrate that the driving force is the stress relaxation at the upper terrace of a diatomic step, which explains the observed nucleation site of semiconductor nanowires.The nanowires investigated consist of InAs͑P͒ selfassembled quantum wires grown by solid source molecular beam epitaxy ͑MBE͒ on InP ͑001͒ substrates. The lattice mismatch between InAs and InP is 3.2%. The control of the relaxation process of these nanostructures has recently been followed with high precision by in situ accumulated stress measurements from the initial phases of the self-assembly process. 23 The process of format...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.