Search

Convolutional Neural Networks (CNNs) are widely deployed in various embedded recognition applications. However, they demonstrate a considerable vulnerability to adversarial attacks, which leverage the well-designed perturbations to mislead the recognition results. Recently, for easier perturbation injection and higher attack effectiveness, the adversarial perturbations have been concentrated into a small area with various types and different data modalities. When defending such condensed adversarial attacks on the embedded recognition scenarios, most of the existing defense works highlight two critical issues. First, they are particularly designed for each individual condensed attack scenario, lacking enough versatility to accommodate attacks with different data modalities. Second, they rely on computation-intensive preprocessing techniques, which is impractical for time-sensitive embedded recognition scenarios. In this article, we propose LanCeX –a versatile and lightweight CNN defense solution against condensed adversarial attacks. By examining the CNN’s intrinsic vulnerability, we first identify the common attacking mechanism behind condensed adversarial attacks across different data modalities. Based on this mechanism, LanCeX can defend against various condensed attacks with the optimal computation workload in different recognition scenarios. Experiments show that LanCeX can achieve an average 91%, 85%, and 90% detection success rate and optimal adversarial mitigation performance in three recognition scenarios, respectively: image classification, object detection, and audio recognition. Moreover, LanCeX is at most 3× faster compared with the state-of-the-art defense methods, making it feasible to use with resource-constrained embedded systems.

Although convolutional neural networks (CNNs) have been widely applied in various cognitive applications, they are still very computationally intensive for resource-constrained mobile systems. To reduce the resource consumption of CNN computation, many optimization works have been proposed for mobile CNN deployment. However, most works are merely targeting CNN model compression from the perspective of parameter size or model structure, ignoring different resource constraints in mobile systems with respect to memory, energy, and real-time requirement. Moreover, previous works take accuracy as their primary consideration, requiring a time-costing retraining process to compensate the inference accuracy loss after compression. To address these issues, we propose DiReCtX —a dynamic resource-aware CNN model reconfiguration framework. DiReCtX is based on a set of accurate CNN profiling models for different resource consumption and inference accuracy estimation. With manageable consumption/accuracy tradeoffs, DiReCtX can reconfigure a CNN model to meet distinct resource constraint types and levels with expected inference performance maintained. To further achieve fast model reconfiguration in real-time, improved CNN model pruning and its corresponding accuracy tuning strategies are also proposed in DiReCtX . The experiments show that the proposed CNN profiling models can achieve 94.6% and 97.1% accuracy for CNN model resource consumption and inference accuracy estimation. Meanwhile, the proposed reconfiguration scheme of DiReCtX can achieve at most 44.44% computation acceleration, 31.69% memory reduction, and 32.39% energy saving, respectively. On field-tests with state-of-the-art smartphones, DiReCtX can adapt CNN models to various resource constraints in mobile application scenarios with optimal real-time performance.

Federated learning learns from scattered data by fusing collaborative models from local nodes. However, the conventional coordinate-based model averaging by FedAvg ignored the random information encoded per parameter and may suffer from structural feature misalignment. In this work, we propose Fed2, a feature-aligned federated learning framework to resolve this issue by establishing a firm structure-feature alignment across the collaborative models. Fed2 is composed of two major designs: First, we design a feature-oriented model structure adaptation method to ensure explicit feature allocation in different neural network structures. Applying the structure adaptation to collaborative models, matchable structures with similar feature information can be initialized at the very early training stage. During the federated learning process, we then propose a feature paired averaging scheme to guarantee aligned feature distribution and maintain no feature fusion conflicts under either IID or non-IID scenarios. Eventually, Fed2 could effectively enhance the federated learning convergence performance under extensive homo- and heterogeneous settings, providing excellent convergence speed, accuracy, and computation/communication efficiency., Accepted in KDD 2021

Latent thermal energy storage has been recommended as an effective technique to the thermal management system of space exploration due to its excellent ability of storing thermal energy. In the current work, the melting process of phase change material under supergravity conditions (3g, 5g, 7g) are numerically investigated by a pore-scale lattice Boltzmann method with the real microstructure of metal foam. A comparative study is carried out to illustrate the influence of gravitational acceleration and metal foam on melting performance of phase change material. Results indicate that natural convection accelerates the motion velocity of the upper PCM, and leads to the inhomogeneous distribution of melting front. Moreover, with the increasing gravitational acceleration, there is a transition of the dominated heat transfer mechanism from conduction to convection, which attributes to the stepped-up natural convection effect, and thus melting front gets even steeper and melting rate is significantly increased. Besides, the existence of metal foam significantly distorts the melting front and enhances the melting rate.

Cement-emulsified asphalt mixture (CEAM), a kind of cold mix asphalt mixture, has the advantages of energy conservation and emission reduction as well as easy construction. However, the performance of CEAM is not as good as hot mixed asphalt mixtures. Hence, in this study, two different fibers were adopted as the reinforcing phase to improve the comprehensive properties of CEAM. The results indicated that the addition proportion and curing time were crucial to fiber-reinforced cement-emulsified asphalt mixture (FRCEAM). The compressive strengths, water stability, and raveling resistances of FRCEAM preparations with polyester or brucite fibers (FRCEAM-PF and -BF, respectively) were enhanced significantly. FRCEAM-PF had the maximum flexural tensile strength and strain, which meant that its low-temperature performance was the best compared to FRCEAM-PF and CEAM. However, the contribution of PF to CEAM high-temperature stability was greater than that of BF. Fiber addition to CEAM not only enhanced the cycles of fatigue loading but also reduced sensitivity to changes in stress level. Furthermore, FRCEAM-BF durability was slightly better than that of FRCEAM-PF. SEM analysis indicated that fibers provided bridging and meshing effects. Although PF and BF showed different enhancement effects, both mixtures met the requirements for hot mixed asphalt mixtures.

We study the ternary Ohta-Kawasaki free energy that has been used to model triblock copolymer systems. Its one-dimensional global minimizers are conjectured to have cyclic patterns. However, some physical experiments and computer simulations found triblock copolymers forming noncyclic lamellar patterns. In this work, by comparing the free energies of the cyclic pattern and some noncyclic candidates, we show that the conjecture does not hold for some choices of parameters. Our results suggest that even in one dimension, the global minimizers may take on very different patterns in different parameter regimes. To unify the existing choices of the long range coefficient matrix, we present a reformulation of the long range term using a generalized charge interpretation, and thereby propose conditions on the matrix in order for the global minimizers to reproduce physically relevant nanostructures of block copolymers., Comment: 35 pages; 12 figures; appeared at AMS 2021 Fall Southeastern Sectional Meeting; minor typos corrected; several references added

Acanthus ilicifolius, also referred to as holly mangrove, usually grows in the coastal areas in southern China. It is not only a pioneer species of mangroves but also an important medicinal mangrove plant. In June 2019, leaf spots on A. ilicifolius with disease incidence of about 30 to 40% were observed in about 150 plants in a field located in Zhanjiang City, Guangdong Province, China. The disease mainly occurred on the leaf margins or tips. Initial symptoms of the disease were dark brown spots (5 to 9 mm) surrounded by a yellow halo, that expanded irregularly or semicircularly, until finally the leaves turned chlorotic and defoliated. Six samples of symptomatic leaves were excised, surface sterilized with 75% ethanol 30 s, 0.1% HgCl2 30 s, and rinsed three times with sterile water, and cut into small pieces (5 mm × 5 mm), placed on potato dextrose agar (PDA) at 25°C and darkness, and two isolates with different morphological characteristics were obtained after 3 days. For pathogenicity tests of the isolates, wounded and non-wounded leaves were inoculated. Fresh wounds were made with a sterile needle. Mycelial plugs from 8-day-old cultures of two isolates, ALY1 and ALY2, on PDA were inoculated on adaxial surface of wounded and non-wounded healthy young leaves of the A. ilicifolius seedlings (2 years old, 4 leaves per plant, 2 replicate plants per isolate), covered with wet cotton and kept in moist chambers at 25 ± 1°C, 80% relative humidity. Leaves inoculated with sterile PDA plugs were used as the control. After 3-5 days, the wounded and non-wounded leaves inoculated with ALY1 showed symptoms of leaf spots, similar to those observed in the field, while leaves inoculated with ALY2 and the control leaves remained symptomless. The pathogenicity tests were repeated three times under the same conditions and similar results were observed. ALY1 was confirmed as the pathogen causing the leaf spot. On PDA, ALY1 produced a gray to grayish brown colony. The conidia were obclavate to cylindrical, straight to slightly curved, pale brown, with 6 to 18 pseudosepta, 62.5 to 225.0×3.8 to 14.8 μm (n=17). According to the morphological characteristics, ALY1 was preliminarily identified as Corynespora cassiicola (Zhang et al.2018). For molecular identification, the internal transcribed spacer (ITS), the actin gene (ACT), the β-tubulin gene (TUB2) and translation elongation factor 1-alpha gene (TEF1-α) of the three single-spore isolates from ALY1 (ALY1-1 to ALY1-3) were amplified and sequenced with primers ITS4/ITS5 (White et al.1990), ACT512F/ACT783R(Carbone and Kohn 1999), T1/T2 (Glass and Donaldson 1995), and EF1-728F/EF1-986R (O'Donnell et al.1998) respectively. The ITS region of ALY1 (-1, -2, -3) (GenBank accession number MN860006, -07, -08) was 99-100% identical (579/585, 586/586, 581/584 base pairs) to a corresponding ITS (MH255527) of C. cassiicola. Similarly, the ACT (MN887504, -05, -06), TUB2 (MN887507, -08, -09) and TEF1-α (MN887501, -02, -03) were 100% (342/342, 342/342, 342/342 base pairs), 100% (664/664, 662/662, 664/664 base pairs) and 100% (290/290, 290/290, 290/290 base pairs) to corresponding genes FJ853033, MH763700, and MK589886 sequences of C. cassiicola in GenBank, respectively. Phylogenetic tree generated with MEGA7 by the neighbor-joining analysis revealed that the isolate's ITS, ACT, TUB2 and TEF1-α sequences were grouped in the same clade to C. cassiicola. Based on both the morphological characteristics and sequences analyses, the pathogen causing leaf spot of A. ilicifolius was identified as C. cassiicola. This species had been reported on the A. ilicifolius in Guam America and Hong Kong China (Lu et al. 2000; Zhuang 2001; Dixon et al. 2009; Sumabat et al. 2018). The best of our knowledge, it is the first report of C. cassiicola causing leaf spot of A. ilicifolius in China. This disease may a potential threat to A. ilicifolius along the coastal regions of southern China.

This article discusses the feasibility of establishing a regional library resource sharing platform by categorizing and investigating the literature resources of various universities in Wuhan. The results indicate that the resources of college libraries at different levels are unevenly distributed, especially the resources of college libraries are obviously insufficient; the resources of the peer college libraries have their own advantages. Constructing a regional university library resource sharing platform can compensate for the shortage of resources in some college libraries, complement each other's library resources, increase the utilization rate of library resources in colleges and universities, and serve the teaching and scientific research.

Recuperator and precooler are key components of supercritical CO2 Brayton cycle. One of promising candidates is printed circuit heat exchanger (PCHE) due to its high compactness and heat transfer performance. However, both of them operate at high pressure, which is a great challenge for their safety. Therefore, accurate and quick mechanical stress analysis is important for PCHE design. In this work, a simplified mechanical stress method based on homogenization method is proposed for PCHE. An optimized rectangular channel, which has the same mechanical characteristics with a semicircular channel is proposed. Then the PCHE numerical model is constructed, and the mechanical stress of PCHEs with semicircular and rectangular channels is examined by elastic stress analysis. The results show that the maximum stresses of the semicircular channel, simplified rectangular channel and optimized rectangular channel are 151.6 MPa, 253.0 MPa and 184.3 MPa, respectively. The mechanical characteristics of the optimized rectangular channel are close to the corresponding semicircular channel, which can improve the design accuracy and is benefit for the weight and volume reduction of PCHE.

Recently, adversarial attacks can be applied to the physical world, causing practical issues to various Convolutional Neural Networks (CNNs) powered applications. Most existing physical adversarial attack defense works only focus on eliminating explicit perturbation patterns from inputs, ignoring interpretation to CNN's intrinsic vulnerability. Therefore, they lack the expected versatility to different attacks and thereby depend on considerable data processing costs. In this paper, we propose LanCe -- a comprehensive and lightweight CNN defense methodology against different physical adversarial attacks. By interpreting CNN's vulnerability, we find that non-semantic adversarial perturbations can activate CNN with significantly abnormal activations and even overwhelm other semantic input patterns' activations. We improve the CNN recognition process by adding a self-verification stage to detect the potential adversarial input with only one CNN inference cost. Based on the detection result, we further propose a data recovery methodology to defend the physical adversarial attacks. We apply such defense methodology into both image and audio CNN recognition scenarios and analyze the computational complexity for each scenario, respectively. Experiments show that our methodology can achieve an average 91% successful rate for attack detection and 89% accuracy recovery. Moreover, it is at most 3x faster compared with the state-of-the-art defense methods, making it feasible to resource-constrained embedded systems, such as mobile devices., 6 pages, 8 figures. arXiv admin note: substantial text overlap with arXiv:1905.08790

A phase field model with two phase fields, representing the concentration and the head-tail separation of amphiphilic molecules, respectively, has been constructed using an extension of the Ohta-Kawasaki model (Macromolecules 19, 2621-2632 (1986)). It is shown that this molecularly-informed phase field model is capable of producing various self-assembled amphiphilic aggregates, such as bilayers, vesicles and micelles. Furthermore, pathways connecting two opposed bilayers with a fusion pore are obtained by using a combination of the phase field model and the string method. Multiple fusion pathways, including a classical pathway and a leaky pathway, have been obtained depending on the initial separation of the two bilayers. The study shed light to the understanding of membrane fusion pathways and, more importantly, laid a foundation for further investigation of more complex membrane morphologies and transitions., 10 pages, 7 figures

Benefited from recent artificial intelligence evolution, Automatic Speech Recognition(ASR) technology has achieved enormous performance improvement and wider application. Unfortunately, ASR is also heavily leveraged by speech eavesdropping, where ASR is used to translate large volume of intercepted vocal speech into text content, causing considerable information leakage. In this work, we propose MASKER– a mobile security enhancement solution to protect the mobile speech data from ASR in eavesdropping. By identifying ASR models’ ubiquitous vulnerability, MASKER is designed to generate human imperceptible adversarial noises into the real-time speech on the mobile device (e.g. phone call and voice message). Even the speech data is exposed to eavesdropping during data transmission, the adversarial noises can effectively perturb the ASR process with significant Word Error Rate (WER). Meanwhile,MASKER is further optimized for mobile user perception quality and enhanced for environmental noises adaptation. Moreover,MASKER has outstanding computation efficiency for mobile system integration. Experiments show that, MASKER can achieve security enhancement with an average WER of 84.55% for ASR perturbation, 32% noise reduction for user perception quality and $16 \times $ faster processing speed compared to the state-of-the-art method.

Recently, the machine learning technologies have been widely used in cognitive applications such as Automatic Speech Recognition (ASR) and Image Recognition (IR). Unfortunately, these techniques have been massively used in unauthorized audio/image data analysis, causing serious privacy leakage. To address this issue, we propose HAMPER in this work, which is a data encryption framework that protects the audio/image data from unauthorized ASR/IR analysis. Leveraging machine learning models' vulnerability to adversarial examples, HAMPER encrypt the audio/image data with adversarial noises to perturb the recognition results of ASR/IR systems. To deploy the proposed framework in extensive platforms (e.g. mobile devices), HAMPER also take into consideration of computation efficiency, perturbation transferability, as well as data attribute configuration. Therefore, rather than focusing on the high-level machine learning models, HAMPER generates adversarial examples from the low-level features. Taking advantage of the light computation load, fundamental impact, and direct configurability of the low-level features, the generated adversarial examples can efficiently and effectively affect the whole ASR/IR systems. Experiment results show that, HAMPER can effectively perturb the unauthorized ASR/IR analysis with 85% Word-Error-Rate (WER) and 83% Image-Error-Rate (IER) respectively. Also, HAMPER achieves faster processing speed with 1.5X speedup for image encryption and even 26X in audio, comparing to the state-of-the-art methods. Moreover, HAMPER achieves strong transferability and configures adversarial examples with desired attributes for better scenario adaptation.

Based on the popular Caputo fractional derivative of order $\beta$ in $(0,1)$, we define the censored fractional derivative on the positive half-line $\mathbb R_+$. This derivative proves to be the Feller generator of the censored (or resurrected) decreasing $\beta$-stable process in $\mathbb R_+$. We provide a series representation for the inverse of this censored fractional derivative, which we use to study general censored initial value problems. We are then able to prove that this censored process hits the boundary in a finite time $\tau_\infty$, whose expectation is proportional to that of the first passage time of the $\beta$-stable subordinator. We also show that the censored relaxation equation is solved by the Laplace transform of $\tau_\infty$. This relaxation solution proves to be a completely monotone series, with algebraic decay one order faster than its Caputo counterpart, leading, surprisingly, to a new regime of fractional relaxation models. Lastly, we discuss how this work identifies a new sub-diffusion model., Comment: 47 pages; minor typos corrected; publisher's note added

MicroRNAs (miRNA) play a key roles in ultraviolet-B (UV-B) induced accumulation of bioactive ingredients in plants. Chrysanthemum (Chrysanthemum morifolium Ramat) is a popular ornamental herb with numerous biological activities, but its miRNAs and their roles in UV-B-induced active ingredient accumulation remain unexplored. High-throughput miRNA sequencing technology was used to identify miRNAs responsive to UV-B radiation. Four libraries were constructed from Chrysanthemum ‘Xiaoyangju’ leaves harvested at different time points under UV-B radiation. In total, 245 miRNAs belonging to 42 classical families were identified. Under UV-B treatment, 137 miRNAs showed differential expression, including 71 known and 67 novel miRNAs,. In addition, 1,779 unique target genes involving diverse biological processes were predicted for 80 known and 85 novel miRNAs. Three key glycolysis genes, Glyceraldehyde-3-Phosphate Dehydrogenase, Pyruvate Kinase and Glucan Endo-1,3-beta-D-Glucosidase, were predicated as potential target genes of miR396f-5p, PC-5p-294053_21 and miR397a, respectively. Among identified target genes, five key genes involved in the biosynthesis of chlorogenic acids and flavonoids were predicated as potential targets genes of MIR4367-p5, PC-3p-40855_284, MIR393b-p5, and PC-3p-114893_93. The expression of identified miRNAs relating to bioactive ingredient biosynthesis was significantly affected by UV-B radiation, suggesting a strong effect of UV-B radiation on bioactive ingredient accumulation in Chrysanthemum. Interestingly, several miRNAs related to flavonoids biosynthesis such as PC-3p-114893_93 and MIR4367-p5, were inhibited by short-term UV-B radiation, suggesting that the contribution of UV-B to flavonoids accumulation depends on a low dose of UV-B radiation.

Phase change thermal storage technology has received considerable attention in the thermal management of electronic devices. Unfortunately, the efficiency using paraffin is too low to meet the heat dissipation requirements of high power devices. In this work, gallium is selected as the phase change material, and the melting mechanism of gallium heat sink in spatial distributions is examined by experiment. Results show that the heat transfer efficiency of regions in the Y-cross section near the heating surface and in the X-cross section near the upper surface of the gallium heat sink are better than other regions. The average effective thermal conductivity of the gallium heat sink reaches 71.97 W/(m·K). The natural convection promotes the overall melting rate, but it results in the inefficient heat transfer in the middle melting stage and worse heat transfer in the later melting stage. Compared with the copper foam/paraffin composite heat sink in the same size, the overall and most of melting time of the gallium heat sink is reduced by 55.5% and 56.0%, respectively. At the same thermal storage capacity, the volume and weight of the gallium heat sink are reduced by 67% and 18%, respectively, compared with the copper foam/paraffin composite heat sink.