Xiang Zhang

Bio: Xiang Zhang is an academic researcher from Baylor College of Medicine. The author has contributed to research in topics: Medicine & Computer science. The author has an hindex of 154, co-authored 1733 publications receiving 117576 citations. Previous affiliations of Xiang Zhang include University of California, Berkeley & University of Texas MD Anderson Cancer Center.

Two local vibrational modes related to hydrogen in GaN

Abstract: We have observed two absorption bands located at around 1730 and 2960 cm−1 in the infrared (IR) absorption spectra from undoped GaN samples which are grown using low pressure metalorganic vapor phase epitaxy and irradiated by gamma ray and then exposed to a radio frequency hydrogen plasma. Proton implantation followed by gamma-ray irradiation of the GaN samples can also activate the IR band at around 1730 cm−1. Based on the experimental results, we tentatively ascribe the 1730 cm−1 band to the local vibrational modes of Ga–H complexes in the vicinity of N vacancies and the 2960 cm−1 band to those of either N–H complexes in the vicinity of Ga vacancies or C–H complexes.

Semiclassical model of stimulated Raman scattering in photonic crystals.

Abstract: We study the stimulated Raman scattering (SRS) of light from an atomic system embedded in a photonic crystal and coherently pumped by a laser field. In our study, the electromagnetic field is treated classically and the atomic system is described quantum mechanically. Considering a decomposition of the pump and Stokes fields into the Bloch modes of the photonic crystals and using a multiscale analysis, we derive the Maxwell-Bloch equations for SRS in photonic crystals. These equations contain effective parameters that characterize the SRS gain, the nonlinear atomic response to the electromagnetic field, and the group velocity and that can be calculated in terms of the Bloch modes of the unperturbed photonic crystal. We show that if the pump laser frequency is tuned near a photonic band edge and the atomic system is carefully chosen such that the Stokes mode matches another photonic band edge, low-threshold, enhanced Raman amplification is possible. Possible physical realizations of SRS in photonic crystals are also discussed.

DeepKey: A Multimodal Biometric Authentication System via Deep Decoding Gaits and Brainwaves

Abstract: Biometric authentication involves various technologies to identify individuals by exploiting their unique, measurable physiological and behavioral characteristics. However, traditional biometric authentication systems (e.g., face recognition, iris, retina, voice, and fingerprint) are at increasing risks of being tricked by biometric tools such as anti-surveillance masks, contact lenses, vocoder, or fingerprint films. In this article, we design a multimodal biometric authentication system named DeepKey, which uses both Electroencephalography (EEG) and gait signals to better protect against such risk. DeepKey consists of two key components: an Invalid ID Filter Model to block unauthorized subjects, and an identification model based on attention-based Recurrent Neural Network (RNN) to identify a subject’s EEG IDs and gait IDs in parallel. The subject can only be granted access while all the components produce consistent affirmations to match the user’s proclaimed identity. We implement DeepKey with a live deployment in our university and conduct extensive empirical experiments to study its technical feasibility in practice. DeepKey achieves the False Acceptance Rate (FAR) and the False Rejection Rate (FRR) of 0 and 1.0%, respectively. The preliminary results demonstrate that DeepKey is feasible, shows consistent superior performance compared to a set of methods, and has the potential to be applied to the authentication deployment in real-world settings.

Exceptional mechanical properties of aluminum matrix composites with heterogeneous structure induced by in-situ graphene nanosheet-Cu hybrids

Abstract: Designing heterogeneous structures is a promising pathway for overcoming the trade-off between strength and toughness in metal matrix composites (MMCs). Herein, we report an innovative strategy for fabricating graphene nanosheet-Cu reinforced Al matrix (GNS-Cu/Al) composites with heterogeneous structure. This strategy involves the consolidation of unique composite powders with core-shell grain structure, which are synthesized with the aid of in-situ GNS-Cu hybrids. Results reveal that the fabricated GNS-Cu/Al composite exhibits multiple microstructural heterogeneities, including both heterogeneous grain structure and reinforcement spatial distribution, which endow the composite with a prominent combination of tensile strength of ∼437 MPa, fracture elongation of ∼12.5% and toughness of ∼48.7 MJ m−3. It is confirmed that such microstructural heterogeneities in GNS-Cu/Al composite contribute significant hetero-deformation induced (HDI) stress strengthening and sustained strain hardening, making the key mechanical properties of GNS-Cu/Al considerably outperform the counterpart of Cu/Al composite. Moreover, the coordinated deformation and crack bridging/blunting behaviors are demonstrated to be responsible for the exceptional toughness of GNS-Cu/Al composite. This work offers a promising bottom-up tactic to fabricate Al matrix composites with heterogeneous structures and superior mechanical performances for structural applications.

Lattice locations of silicon atoms in δ-doped layers in GaAs at high doping concentrations

Abstract: Low-noise infrared (IR) absorption measurements of localized vibrational modes (LVM's) showed that ${\mathrm{Si}}_{\mathrm{As}}$ acceptors, ${\mathrm{Si}}_{\mathrm{Ga}}$-${\mathrm{Si}}_{\mathrm{As}}$ pairs, and a deep trap Si-X (${\mathit{V}}_{\mathrm{Ga}}$-${\mathrm{Si}}_{\mathrm{As}}$-${\mathrm{As}}_{\mathrm{Ga}}$), as well as isolated ${\mathrm{Si}}_{\mathrm{Ga}}$ donors, were present in silicon \ensuremath{\delta}-doping superlattices in (001) GaAs grown under an As flux by molecular-beam epitaxy (MBE) at 400 \ifmmode^\circ\else\textdegree\fi{}C for areal concentrations (per layer) 0.05 ML\ensuremath{\leqslant} [Si${]}_{\mathit{A}}$\ensuremath{\leqslant}0.5 ML. These observations supersede previous data, and agree with recent Raman-scattering measurements. For [Si${]}_{\mathit{A}}$\ensuremath{\geqslant}0.5 ML, the LVM's were not detected by either technique, but Raman measurements revealed a broad line that has been attributed to small two-dimensional Si clusters. For [Si${]}_{\mathit{A}}$\ensuremath{\geqslant}0.5 ML, electrical conductivity was lost. These observations led to a reappraisal of simulations of high-resolution x-ray 002 and 004 diffraction profiles. IR and Raman measurements for \ensuremath{\delta}-doping superlattices that all have [Si${]}_{\mathit{A}}$=0.01 ML (per layer) showed only the ${\mathrm{Si}}_{\mathrm{Ga}}$ LVM as the interlayer spacing was reduced to 5 ML when the volume carrier concentration n approached \ensuremath{\sim}2\ifmmode\times\else\texttimes\fi{}${10}^{19}$ ${\mathrm{cm}}^{\mathrm{\ensuremath{-}}3}$. For interlayer spacings of 2 and 1 ML, compensating complexes ${\mathrm{Si}}_{\mathrm{As}}$, ${\mathrm{Si}}_{\mathrm{Ga}}$-${\mathrm{Si}}_{\mathrm{As}}$, and Si-X were present, and n tended to zero. Compensating complexes were also present in homogeneously doped MBE GaAs grown at 350 \ifmmode^\circ\else\textdegree\fi{}C, but n remained at a value of 2\ifmmode\times\else\texttimes\fi{}${10}^{19}$ ${\mathrm{cm}}^{\mathrm{\ensuremath{-}}3}$ as [Si] was increased to 1.3\ifmmode\times\else\texttimes\fi{}${10}^{20}$ ${\mathrm{cm}}^{\mathrm{\ensuremath{-}}3}$. N never exceeded 2\ifmmode\times\else\texttimes\fi{}${10}^{19}$ ${\mathrm{cm}}^{\mathrm{\ensuremath{-}}3}$ in any sample. The formation of ${\mathit{V}}_{\mathrm{Ga}}$, ${\mathrm{As}}_{\mathrm{Ga}}$, etc. is attributed to diffusion jumps of Si atoms originally located on Ga lattice sites. The formation of the ``Si-like'' structure in \ensuremath{\delta} layers must result from the aggregation of such displaced atoms. We speculate that these processes are facilitated by the initial displacements of ${\mathrm{Si}}_{\mathrm{Ga}}$ donors to DX locations. \textcopyright{} 1996 The American Physical Society.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.