Specific features of transmutational doping of 30Si-enriched silicon crystals with phosphorus: Studies by the method of electron spin resonance

Electron spin resonance (ESR) is used to study the neutron transmutation doping of silicon crystals enriched with ³⁰Si isotope: phosphorus donors and radiation defects produced in the course of transmutational doping are observed. The ESR signals related to the phosphorus uncontrolled impurity in ³⁰Si before transmutational doping (the P concentration is ～10¹⁵ cm^?3) and phosphorus introduced by neutron irradiation with doses ～1 × 10¹⁹ cm^?2 and ～1 × 10²⁰ cm^?2 (the P concentrations are ～5 × 10¹⁶ and ～7 × 10¹⁷ cm^?3, respectively) are studied. As a result of drastic narrowing of the phosphorus ESR lines in ³⁰Si, the intensity of lines increased appreciably, which made it possible to measure the phosphorus concentration in the samples with a small volume (down to 10^?6 mm^?3). The methods for determining the concentration of P donors from hyperfine structure in the ESR spectra of isolated P atoms, exchange-related pairs, and clusters that consist of three, four, and more P donors are developed. In the region of high concentrations of P donors, in which case the hyperfine structure disappears, the concentration of P donors was estimated from the exchange-narrowed ESR line.     

Supervised Learning of Quantizer Codebooks by Information Loss Minimization

This paper proposes a technique for jointly quantizing continuous features and the posterior distributions of their class labels based on minimizing empirical information loss such that the quantizer index of a given feature vector approximates a sufficient statistic for its class label. Informally, the quantized representation retains as much information as possible for classifying the feature vector correctly. We derive an alternating minimization procedure for simultaneously learning codebooks in the euclidean feature space and in the simplex of posterior class distributions. The resulting quantizer can be used to encode unlabeled points outside the training set and to predict their posterior class distributions, and has an elegant interpretation in terms of lossless source coding. The proposed method is validated on synthetic and real data sets and is applied to two diverse problems: learning discriminative visual vocabularies for bag-of-features image classification and image segmentation.     

Highly Accurate Debye Models for Normal and Malignant Breast Tissue Dielectric Properties at Microwave Frequencies

The finite difference time domain (FDTD) method is widely used as a computational tool for development, validation, and optimization of emerging microwave breast cancer detection and treatment techniques. When expressed in terms of Debye parameters, dispersive breast tissue dielectric properties can be efficiently incorporated into FDTD codes. Previously, we experimentally characterized the dielectric properties of a large number of excised normal and malignant breast tissue samples from 0.5 to 20 GHz. We subdivided the large database of normal tissue data into three groups based on the percent adipose tissue present in a particular sample. In addition, we formed a group of all cancer samples that contained at least 30% malignant tissue. We summarized the data using one-pole Cole-Cole models that were rigorously fit to the median dielectric properties of the three normal tissue groups and one malignant tissue group. In this letter, we present computationally simpler one- and two-pole Debye models that retain the high accuracy of the Cole-Cole models. Model parameters are derived for two sets of frequency ranges: the entire measurement frequency range from 0.5 to 20 GHz, and the 3.1-10.6 GHz FCC band allocated for ultrawideband medical applications. The proposed Debye models provide a means for creating computationally efficient FDTD breast models with realistic wideband dielectric properties derived from the largest and most comprehensive experimental study conducted to date on human breast tissue.     

Combining Multiple Cues for Visual Madlibs Question Answering

International Journal of Computer Vision - This paper presents an approach for answering fill-in-the-blank multiple choice questions from the Visual Madlibs dataset. Instead of generic and commonly...