Showing papers by "University of Central Florida published in 2017"

3D printing of polymer matrix composites: A review and prospective

Abstract: The use of 3D printing for rapid tooling and manufacturing has promised to produce components with complex geometries according to computer designs. Due to the intrinsically limited mechanical properties and functionalities of printed pure polymer parts, there is a critical need to develop printable polymer composites with high performance. 3D printing offers many advantages in the fabrication of composites, including high precision, cost effective and customized geometry. This article gives an overview on 3D printing techniques of polymer composite materials and the properties and performance of 3D printed composite parts as well as their potential applications in the fields of biomedical, electronics and aerospace engineering. Common 3D printing techniques such as fused deposition modeling, selective laser sintering, inkjet 3D printing, stereolithography, and 3D plotting are introduced. The formation methodology and the performance of particle-, fiber- and nanomaterial-reinforced polymer composites are emphasized. Finally, important limitations are identified to motivate the future research of 3D printing.

Enhanced sensitivity at higher-order exceptional points

Abstract: Non-Hermitian degeneracies, also known as exceptional points, have recently emerged as a new way to engineer the response of open physical systems, that is, those that interact with the environment. They correspond to points in parameter space at which the eigenvalues of the underlying system and the corresponding eigenvectors simultaneously coalesce. In optics, the abrupt nature of the phase transitions that are encountered around exceptional points has been shown to lead to many intriguing phenomena, such as loss-induced transparency, unidirectional invisibility, band merging, topological chirality and laser mode selectivity. Recently, it has been shown that the bifurcation properties of second-order non-Hermitian degeneracies can provide a means of enhancing the sensitivity (frequency shifts) of resonant optical structures to external perturbations. Of particular interest is the use of even higher-order exceptional points (greater than second order), which in principle could further amplify the effect of perturbations, leading to even greater sensitivity. Although a growing number of theoretical studies have been devoted to such higher-order degeneracies, their experimental demonstration in the optical domain has so far remained elusive. Here we report the observation of higher-order exceptional points in a coupled cavity arrangement-specifically, a ternary, parity-time-symmetric photonic laser molecule-with a carefully tailored gain-loss distribution. We study the system in the spectral domain and find that the frequency response associated with this system follows a cube-root dependence on induced perturbations in the refractive index. Our work paves the way for utilizing non-Hermitian degeneracies in fields including photonics, optomechanics, microwaves and atomic physics.

Asymmetric Supercapacitor Electrodes and Devices.

Abstract: The world is recently witnessing an explosive development of novel electronic and optoelectronic devices that demand more-reliable power sources that combine higher energy density and longer-term durability. Supercapacitors have become one of the most promising energy-storage systems, as they present multifold advantages of high power density, fast charging-discharging, and long cyclic stability. However, the intrinsically low energy density inherent to traditional supercapacitors severely limits their widespread applications, triggering researchers to explore new types of supercapacitors with improved performance. Asymmetric supercapacitors (ASCs) assembled using two dissimilar electrode materials offer a distinct advantage of wide operational voltage window, and thereby significantly enhance the energy density. Recent progress made in the field of ASCs is critically reviewed, with the main focus on an extensive survey of the materials developed for ASC electrodes, as well as covering the progress made in the fabrication of ASC devices over the last few decades. Current challenges and a future outlook of the field of ASCs are also discussed.

Deeply Supervised Salient Object Detection with Short Connections

Abstract: Recent progress on saliency detection is substantial, benefiting mostly from the explosive development of Convolutional Neural Networks (CNNs). Semantic segmentation and saliency detection algorithms developed lately have been mostly based on Fully Convolutional Neural Networks (FCNs). There is still a large room for improvement over the generic FCN models that do not explicitly deal with the scale-space problem. Holisitcally-Nested Edge Detector (HED) provides a skip-layer structure with deep supervision for edge and boundary detection, but the performance gain of HED on saliency detection is not obvious. In this paper, we propose a new saliency method by introducing short connections to the skip-layer structures within the HED architecture. Our framework provides rich multi-scale feature maps at each layer, a property that is critically needed to perform segment detection. Our method produces state-of-the-art results on 5 widely tested salient object detection benchmarks, with advantages in terms of efficiency (0.08 seconds per image), effectiveness, and simplicity over the existing algorithms.

Structure-Measure: A New Way to Evaluate Foreground Maps

Abstract: Foreground map evaluation is crucial for gauging the progress of object segmentation algorithms, in particular in the field of salient object detection where the purpose is to accurately detect and segment the most salient object in a scene. Several widely-used measures such as Area Under the Curve (AUC), Average Precision (AP) and the recently proposed F W/B (Fbw) have been used to evaluate the similarity between a non-binary saliency map (SM) and a ground-truth (GT) map. These measures are based on pixel-wise errors and often ignore the structural similarities. Behavioral vision studies, however, have shown that the human visual system is highly sensitive to structures in scenes. Here, we propose a novel, efficient, and easy to calculate measure known as structural similarity measure (Structure-measure) to evaluate non-binary foreground maps. Our new measure simultaneously evaluates region-aware and object-aware structural similarity between a SM and a GT map. We demonstrate superiority of our measure over existing ones using 5 meta-measures on 5 benchmark datasets.

A Stagewise Refinement Model for Detecting Salient Objects in Images

Abstract: Deep convolutional neural networks (CNNs) have been successfully applied to a wide variety of problems in computer vision, including salient object detection. To detect and segment salient objects accurately, it is necessary to extract and combine high-level semantic features with low-levelfine details simultaneously. This happens to be a challenge for CNNs as repeated subsampling operations such as pooling and convolution lead to a significant decrease in the initial image resolution, which results in loss of spatial details and finer structures. To remedy this problem, here we propose to augment feedforward neural networks with a novel pyramid pooling module and a multi-stage refinement mechanism for saliency detection. First, our deep feedward net is used to generate a coarse prediction map with much detailed structures lost. Then, refinement nets are integrated with local context information to refine the preceding saliency maps generated in the master branch in a stagewise manner. Further, a pyramid pooling module is applied for different-region-based global context aggregation. Empirical evaluations over six benchmark datasets show that our proposed method compares favorably against the state-of-the-art approaches.

Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes

Abstract: During the last half decade, convolutional neural networks (CNNs) have triumphed over semantic segmentation, which is a core task of various emerging industrial applications such as autonomous driving and medical imaging. However, to train CNNs requires a huge amount of data, which is difficult to collect and laborious to annotate. Recent advances in computer graphics make it possible to train CNN models on photo-realistic synthetic data with computer-generated annotations. Despite this, the domain mismatch between the real images and the synthetic data significantly decreases the models’ performance. Hence we propose a curriculum-style learning approach to minimize the domain gap in semantic segmentation. The curriculum domain adaptation solves easy tasks first in order to infer some necessary properties about the target domain; in particular, the first task is to learn global label distributions over images and local distributions over landmark superpixels. These are easy to estimate because images of urban traffic scenes have strong idiosyncrasies (e.g., the size and spatial relations of buildings, streets, cars, etc.). We then train the segmentation network in such a way that the network predictions in the target domain follow those inferred properties. In experiments, our method significantly outperforms the baselines as well as the only known existing approach to the same problem.

Semi Supervised Semantic Segmentation Using Generative Adversarial Network

Abstract: Semantic segmentation has been a long standing challenging task in computer vision. It aims at assigning a label to each image pixel and needs a significant number of pixel-level annotated data, which is often unavailable. To address this lack of annotations, in this paper, we leverage, on one hand, a massive amount of available unlabeled or weakly labeled data, and on the other hand, non-real images created through Generative Adversarial Networks. In particular, we propose a semi-supervised framework – based on Generative Adversarial Networks (GANs) – which consists of a generator network to provide extra training examples to a multi-class classifier, acting as discriminator in the GAN framework, that assigns sample a label y from the K possible classes or marks it as a fake sample (extra class). The underlying idea is that adding large fake visual data forces real samples to be close in the feature space, which, in turn, improves multiclass pixel classification. To ensure a higher quality of generated images by GANs with consequently improved pixel classification, we extend the above framework by adding weakly annotated data, i.e., we provide class level information to the generator. We test our approaches on several challenging benchmarking visual datasets, i.e. PASCAL, SiftFLow, Stanford and CamVid, achieving competitive performance compared to state-of-the-art semantic segmentation methods.

Structure-measure: A New Way to Evaluate Foreground Maps

Abstract: Foreground map evaluation is crucial for gauging the progress of object segmentation algorithms, in particular in the filed of salient object detection where the purpose is to accurately detect and segment the most salient object in a scene. Several widely-used measures such as Area Under the Curve (AUC), Average Precision (AP) and the recently proposed Fbw have been utilized to evaluate the similarity between a non-binary saliency map (SM) and a ground-truth (GT) map. These measures are based on pixel-wise errors and often ignore the structural similarities. Behavioral vision studies, however, have shown that the human visual system is highly sensitive to structures in scenes. Here, we propose a novel, efficient, and easy to calculate measure known an structural similarity measure (Structure-measure) to evaluate non-binary foreground maps. Our new measure simultaneously evaluates region-aware and object-aware structural similarity between a SM and a GT map. We demonstrate superiority of our measure over existing ones using 5 meta-measures on 5 benchmark datasets.

Regeneration of fat cells from myofibroblasts during wound healing

Abstract: Although regeneration through the reprogramming of one cell lineage to another occurs in fish and amphibians, it has not been observed in mammals. We discovered in the mouse that during wound healing, adipocytes regenerate from myofibroblasts, a cell type thought to be differentiated and nonadipogenic. Myofibroblast reprogramming required neogenic hair follicles, which triggered bone morphogenetic protein (BMP) signaling and then activation of adipocyte transcription factors expressed during development. Overexpression of the BMP antagonist Noggin in hair follicles or deletion of the BMP receptor in myofibroblasts prevented adipocyte formation. Adipocytes formed from human keloid fibroblasts either when treated with BMP or when placed with human hair follicles in vitro. Thus, we identify the myofibroblast as a plastic cell type that may be manipulated to treat scars in humans.

Blockchain technology innovations

Abstract: Digital world has produced efficiencies, new innovative products, and close customer relationships globally by the effective use of mobile, IoT (Internet of Things), social media, analytics and cloud technology to generate models for better decisions. Blockchain is recently introduced and revolutionizing the digital world bringing a new perspective to security, resiliency and efficiency of systems. While initially popularized by Bitcoin, Blockchain is much more than a foundation for crypto currency. It offers a secure way to exchange any kind of good, service, or transaction. Industrial growth increasingly depends on trusted partnerships; but increasing regulation, cybercrime and fraud are inhibiting expansion. To address these challenges, Blockchain will enable more agile value chains, faster product innovations, closer customer relationships, and quicker integration with the IoT and cloud technology. Further Blockchain provides a lower cost of trade with a trusted contract monitored without intervention from third parties who may not add direct value. It facilitates smart contracts, engagements, and agreements with inherent, robust cyber security features. This paper is an effort to break the ground for presenting and demonstrating the use of Blockchain technology in multiple industrial applications. A healthcare industry application, Healthchain, is formalized and developed on the foundation of Blockchain using IBM Blockchain initiative. The concepts are transferable to a wide range of industries as finance, government and manufacturing where security, scalability and efficiency must meet.

Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos

Abstract: Deep learning has been demonstrated to achieve excellent results for image classification and object detection. However, the impact of deep learning on video analysis has been limited due to complexity of video data and lack of annotations. Previous convolutional neural networks (CNN) based video action detection approaches usually consist of two major steps: frame-level action proposal generation and association of proposals across frames. Also, most of these methods employ two-stream CNN framework to handle spatial and temporal feature separately. In this paper, we propose an end-to-end deep network called Tube Convolutional Neural Network (T-CNN) for action detection in videos. The proposed architecture is a unified deep network that is able to recognize and localize action based on 3D convolution features. A video is first divided into equal length clips and next for each clip a set of tube proposals are generated based on 3D Convolutional Network (ConvNet) features. Finally, the tube proposals of different clips are linked together employing network flow and spatio-temporal action detection is performed using these linked video proposals. Extensive experiments on several video datasets demonstrate the superior performance of T-CNN for classifying and localizing actions in both trimmed and untrimmed videos compared to state-of-the-arts.

The global distribution of tetrapods reveals a need for targeted reptile conservation

Abstract: The distributions of amphibians, birds and mammals have underpinned global and local conservation priorities, and have been fundamental to our understanding of the determinants of global biodiversity. In contrast, the global distributions of reptiles, representing a third of terrestrial vertebrate diversity, have been unavailable. This prevented the incorporation of reptiles into conservation planning and biased our understanding of the underlying processes governing global vertebrate biodiversity. Here, we present and analyse the global distribution of 10,064 reptile species (99% of extant terrestrial species). We show that richness patterns of the other three tetrapod classes are good spatial surrogates for species richness of all reptiles combined and of snakes, but characterize diversity patterns of lizards and turtles poorly. Hotspots of total and endemic lizard richness overlap very little with those of other taxa. Moreover, existing protected areas, sites of biodiversity significance and global conservation schemes represent birds and mammals better than reptiles. We show that additional conservation actions are needed to effectively protect reptiles, particularly lizards and turtles. Adding reptile knowledge to a global complementarity conservation priority scheme identifies many locations that consequently become important. Notably, investing resources in some of the world’s arid, grassland and savannah habitats might be necessary to represent all terrestrial vertebrates efficiently.

Ultrathin High Surface Area Nickel Boride (NixB) Nanosheets as Highly Efficient Electrocatalyst for Oxygen Evolution

Abstract: The overriding obstacle to mass production of hydrogen from water as the premium fuel for powering our planet is the frustratingly slow kinetics of the oxygen evolution reaction (OER). Additionally, inadequate understanding of the key barriers of the OER is a hindrance to insightful design of advanced OER catalysts. This study presents ultrathin amorphous high-surface area nickel boride (NixB) nanosheets as a low-cost, very efficient and stable catalyst for the OER for electrochemical water splitting. The catalyst affords 10 mA cm−2 at 0.38 V overpotential during OER in 1.0 m KOH, reducing to only 0.28 V at 20 mA cm−2 when supported on nickel foam, which ranks it among the best reported nonprecious catalysts for oxygen evolution. Operando X-ray absorption fine-structure spectroscopy measurements reveal prevalence of NiOOH, as well as Ni-B under OER conditions, owing to a Ni-B core@nickel oxyhydroxide shell (Ni-B@NiOxH) structure, and increase in disorder of the NiOxH layer, thus revealing important insight into the transient states of the catalyst during oxygen evolution.

53-attosecond X-ray pulses reach the carbon K-edge

Abstract: The motion of electrons in the microcosm occurs on a time scale set by the atomic unit of time—24 attoseconds. Attosecond pulses at photon energies corresponding to the fundamental absorption edges of matter, which lie in the soft X-ray regime above 200 eV, permit the probing of electronic excitation, chemical state, and atomic structure. Here we demonstrate a soft X-ray pulse duration of 53 as and single pulse streaking reaching the carbon K-absorption edge (284 eV) by utilizing intense two-cycle driving pulses near 1.8-μm center wavelength. Such pulses permit studies of electron dynamics in live biological samples and next-generation electronic materials such as diamond. Isolated attosecond pulses are produced using high harmonic generation and sources of these pulses often suffer from low photon flux in soft X-ray regime. Here the authors demonstrate efficient generation and characterization of 53 as pulses with photon energy near the water window.

AppSAT: Approximately deobfuscating integrated circuits

Abstract: In today's diversified semiconductor supply-chain, protecting intellectual property (IP) and maintaining manufacturing integrity are important concerns. Circuit obfuscation techniques such as logic encryption and IC camouflaging can potentially defend against a majority of supply-chain threats such as stealthy malicious design modification, IP theft, overproduction, and cloning. Recently, a Boolean Satisfiability (SAT) based attack, namely the SAT attack has been able to deobfuscate almost all traditional circuit obfuscation schemes, and as a result, a number of defense solutions have been proposed in literature. All these defenses are based on the implicit assumption that the attacker needs a perfect deobfuscation accuracy which may not be true in many practical cases. Therefore, in this paper by relaxing the exactness constraint on deobfuscation, we propose the AppSAT attack, an approximate deobfuscation algorithm based on the SAT attack and random testing. We show how the AppSAT attack can deobfuscate 68 out of the 71 benchmark circuits that were obfuscated with state-of-the-art SAT attack defenses with an accuracy of, n being the number of inputs. AppSAT shows that with current SAT attack defenses there will be a trade-off between exact-attack resiliency and approximation resiliency.

Interleaved Group Convolutions

Abstract: In this paper, we present a simple and modularized neural network architecture, named interleaved group convolutional neural networks (IGCNets). The main point lies in a novel building block, a pair of two successive interleaved group convolutions: primary group convolution and secondary group convolution. The two group convolutions are complementary: (i) the convolution on each partition in primary group convolution is a spatial convolution, while on each partition in secondary group convolution, the convolution is a point-wise convolution; (ii) the channels in the same secondary partition come from different primary partitions. We discuss one representative advantage: Wider than a regular convolution with the number of parameters and the computation complexity preserved. We also show that regular convolutions, group convolution with summation fusion, and the Xception block are special cases of interleaved group convolutions. Empirical results over standard benchmarks, CIFAR-10, CIFAR-100, SVHN and ImageNet demonstrate that our networks are more efficient in using parameters and computation complexity with similar or higher accuracy.

Leadership training design, delivery, and implementation: A meta-analysis.

Abstract: Recent estimates suggest that although a majority of funds in organizational training budgets tend to be allocated to leadership training (Ho, 2016; O'Leonard, 2014), only a small minority of organizations believe their leadership training programs are highly effective (Schwartz, Bersin, & Pelster, 2014), calling into question the effectiveness of current leadership development initiatives. To help address this issue, this meta-analysis estimates the extent to which leadership training is effective and identifies the conditions under which these programs are most effective. In doing so, we estimate the effectiveness of leadership training across four criteria (reactions, learning, transfer, and results; Kirkpatrick, 1959) using only employee data and we examine 15 moderators of training design and delivery to determine which elements are associated with the most effective leadership training interventions. Data from 335 independent samples suggest that leadership training is substantially more effective than previously thought, leading to improvements in reactions (δ = .63), learning (δ = .73), transfer (δ = .82), and results (δ = .72), the strength of these effects differs based on various design, delivery, and implementation characteristics. Moderator analyses support the use of needs analysis, feedback, multiple delivery methods (especially practice), spaced training sessions, a location that is on-site, and face-to-face delivery that is not self-administered. Results also suggest that the content of training, attendance policy, and duration influence the effectiveness of the training program. Practical implications for training development and theoretical implications for leadership and training literatures are discussed. (PsycINFO Database Record

Soft robotic sleeve supports heart function

Abstract: There is much interest in form-fitting, low-modulus, implantable devices or soft robots that can mimic or assist in complex biological functions such as the contraction of heart muscle. We present a soft robotic sleeve that is implanted around the heart and actively compresses and twists to act as a cardiac ventricular assist device. The sleeve does not contact blood, obviating the need for anticoagulation therapy or blood thinners, and reduces complications with current ventricular assist devices, such as clotting and infection. Our approach used a biologically inspired design to orient individual contracting elements or actuators in a layered helical and circumferential fashion, mimicking the orientation of the outer two muscle layers of the mammalian heart. The resulting implantable soft robot mimicked the form and function of the native heart, with a stiffness value of the same order of magnitude as that of the heart tissue. We demonstrated feasibility of this soft sleeve device for supporting heart function in a porcine model of acute heart failure. The soft robotic sleeve can be customized to patient-specific needs and may have the potential to act as a bridge to transplant for patients with heart failure.

Stock Price Prediction via Discovering Multi-Frequency Trading Patterns

Abstract: Stock prices are formed based on short and/or long-term commercial and trading activities that reflect different frequencies of trading patterns. However, these patterns are often elusive as they are affected by many uncertain political-economic factors in the real world, such as corporate performances, government policies, and even breaking news circulated across markets. Moreover, time series of stock prices are non-stationary and non-linear, making the prediction of future price trends much challenging. To address them, we propose a novel State Frequency Memory (SFM) recurrent network to capture the multi-frequency trading patterns from past market data to make long and short term predictions over time. Inspired by Discrete Fourier Transform (DFT), the SFM decomposes the hidden states of memory cells into multiple frequency components, each of which models a particular frequency of latent trading pattern underlying the fluctuation of stock price. Then the future stock prices are predicted as a nonlinear mapping of the combination of these components in an Inverse Fourier Transform (IFT) fashion. Modeling multi-frequency trading patterns can enable more accurate predictions for various time ranges: while a short-term prediction usually depends on high frequency trading patterns, a long-term prediction should focus more on the low frequency trading patterns targeting at long-term return. Unfortunately, no existing model explicitly distinguishes between various frequencies of trading patterns to make dynamic predictions in literature. The experiments on the real market data also demonstrate more competitive performance by the SFM as compared with the state-of-the-art methods.

Dynamically Encircling Exceptional Points: Exact Evolution and Polarization State Conversion.

Abstract: We show that a two-level non-Hermitian Hamiltonian with constant off-diagonal exchange elements can be analyzed exactly when the underlying exceptional point is perfectly encircled in the complex plane. The state evolution of this system is explicitly obtained in terms of an ensuing transfer matrix, even for large encirclements, regardless of adiabatic conditions. Our results clearly explain the direction-dependent nature of this process and why in the adiabatic limit its outcome is dominated by a specific eigenstate-irrespective of initial conditions. Moreover, numerical simulations suggest that this mechanism can still persist in the presence of nonlinear effects. We further show that this robust process can be harnessed to realize an optical omnipolarizer: a configuration that generates a desired polarization output regardless of the input polarization state, while from the opposite direction it always produces the counterpart eigenstate.

Fetal Bovine Serum (FBS): Past - Present - Future.

Abstract: The supplementation of culture medium with fetal bovine serum (FBS, also referred to as "fetal calf serum") is still common practice in cell culture applications. Due to a number of disadvantages in terms of quality and reproducibility of in vitro data, animal welfare concerns, and in light of recent cases of fraudulent marketing, the search for alternatives and the development of serum-free medium formulations has gained global attention. Here, we report on the 3rd Workshop on FBS, Serum Alternatives and Serum-free Media, where regulatory aspects, the serum dilemma, alternatives to FBS, case-studies of serum-free in vitro applications, and the establishment of serum-free databases were discussed. The whole process of obtaining blood from a living calf fetus to using the FBS produced from it for scientific purposes is de facto not yet legally regulated despite the existing EU-Directive 2010/63/EU on the use of animals for scientific purposes. Together with the above-mentioned challenges, several strategies have been developed to reduce or replace FBS in cell culture media in terms of the 3Rs (Refinement, Reduction, Replacement). Most recently, releasates of activated human donor thrombocytes (human platelet lysates) have been shown to be one of the most promising serum alternatives when chemically-defined media are not yet an option. Additionally, new developments in cell-based assay techniques, advanced organ-on-chip and microphysiological systems are covered in this report. Chemically-defined serum-free media are shown to be the ultimate goal for the majority of culture systems, and examples are discussed.