arXiv subjects

Xiaowei Li

Publications and source records attributed to Xiaowei Li.

At least 73 records · Page 4Linked to original sources

arXiv2021

Taming Process Variations in CNFET for Efficient Last Level Cache Design

Carbon nanotube field-effect transistors (CNFET) emerge as a promising alternative to CMOS transistors for the much higher speed and energy efficiency, which makes the technology particularly suitable for building the energy-hungry last level cache (LLC). However, the process variations (PVs) in CNFET caused by the imperfect fabrication lead to large timing variation and the worst-case timing dramatically limits the LLC operation speed. Particularly, we observe that the CNFET-based cache latency distribution is closely related to the LLC layouts. For the two typical LLC layouts that have the CNT growth direction aligned to the cache way direction and cache set direction respectively, we proposed variation-aware set aligned (VASA) cache and variation-aware way aligned (VAWA) cache in combination with corresponding cache optimizations such as data shuffling and page mapping to enable low-latency cache for frequently used data. According to our experiments, the optimized LLC reduces the average access latency by 32% and 45% compared to the baseline designs on the two different CNFET layouts respectively while it improves the overall performance by 6\% and 9\% and reduces the energy consumption by 4% and 8% respectively. In addition, with both the architecture induced latency variation and PV incurred latency variation considered in a unified model, we extended the VAWA and VASA cache design for the CNFET-based NUCA and the proposed NUCA achieves both significant performance improvement and energy saving compared to the straightforward variation-aware NUCA.

cs.AR↗

arXiv2021

R2F: A Remote Retraining Framework for AIoT Processors with Computing Errors

AIoT processors fabricated with newer technology nodes suffer rising soft errors due to the shrinking transistor sizes and lower power supply. Soft errors on the AIoT processors particularly the deep learning accelerators (DLAs) with massive computing may cause substantial computing errors. These computing errors are difficult to be captured by the conventional training on general purposed processors like CPUs and GPUs in a server. Applying the offline trained neural network models to the edge accelerators with errors directly may lead to considerable prediction accuracy loss. To address the problem, we propose a remote retraining framework (R2F) for remote AIoT processors with computing errors. It takes the remote AIoT processor with soft errors in the training loop such that the on-site computing errors can be learned with the application data on the server and the retrained models can be resilient to the soft errors. Meanwhile, we propose an optimized partial TMR strategy to enhance the retraining. According to our experiments, R2F enables elastic design trade-offs between the model accuracy and the performance penalty. The top-5 model accuracy can be improved by 1.93%-13.73% with 0%-200% performance penalty at high fault error rate. In addition, we notice that the retraining requires massive data transmission and even dominates the training time, and propose a sparse increment compression approach for the data transmission optimization, which reduces the retraining time by 38%-88% on average with negligible accuracy loss over a straightforward remote retraining.

cs.AR↗

arXiv2021

Energy-Efficient Accelerator Design for Deformable Convolution Networks

Deformable convolution networks (DCNs) proposed to address the image recognition with geometric or photometric variations typically involve deformable convolution that convolves on arbitrary locations of input features. The locations change with different inputs and induce considerable dynamic and irregular memory accesses which cannot be handled by classic neural network accelerators (NNAs). Moreover, bilinear interpolation (BLI) operation that is required to obtain deformed features in DCNs also cannot be deployed on existing NNAs directly. Although a general purposed processor (GPP) seated along with classic NNAs can process the deformable convolution, the processing on GPP can be extremely slow due to the lack of parallel computing capability. To address the problem, we develop a DCN accelerator on existing NNAs to support both the standard convolution and deformable convolution. Specifically, for the dynamic and irregular accesses in DCNs, we have both the input and output features divided into tiles and build a tile dependency table (TDT) to track the irregular tile dependency at runtime. With the TDT, we further develop an on-chip tile scheduler to handle the dynamic and irregular accesses efficiently. In addition, we propose a novel mapping strategy to enable parallel BLI processing on NNAs and apply layer fusion techniques for more energy-efficient DCN processing. According to our experiments, the proposed accelerator achieves orders of magnitude higher performance and energy efficiency compared to the typical computing architectures including ARM, ARM+TPU, and GPU with 6.6\% chip area penalty to a classic NNA.

cs.AR↗

arXiv2021

Sub-Gaussian Matrices on Sets: Optimal Tail Dependence and Applications

Random linear mappings are widely used in modern signal processing, compressed sensing and machine learning. These mappings may be used to embed the data into a significantly lower dimension while at the same time preserving useful information. This is done by approximately preserving the distances between data points, which are assumed to belong to $\mathbb{R}^n$. Thus, the performance of these mappings is usually captured by how close they are to an isometry on the data. Gaussian linear mappings have been the object of much study, while the sub-Gaussian settings is not yet fully understood. In the latter case, the performance depends on the sub-Gaussian norm of the rows. In many applications, e.g., compressed sensing, this norm may be large, or even growing with dimension, and thus it is important to characterize this dependence. We study when a sub-Gaussian matrix can become a near isometry on a set, show that previous best known dependence on the sub-Gaussian norm was sub-optimal, and present the optimal dependence. Our result not only answers a remaining question posed by Liaw, Mehrabian, Plan and Vershynin in 2017, but also generalizes their work. We also develop a new Bernstein type inequality for sub-exponential random variables, and a new Hanson-Wright inequality for quadratic forms of sub-Gaussian random variables, in both cases improving the bounds in the sub-Gaussian regime under moment constraints. Finally, we illustrate popular applications such as Johnson-Lindenstrauss embeddings, null space property for 0-1 matrices, randomized sketches and blind demodulation, whose theoretical guarantees can be improved by our results (in the sub-Gaussian case).

cs.IT↗

arXiv2020

Exploring Spatial-Temporal Multi-Frequency Analysis for High-Fidelity and Temporal-Consistency Video Prediction

Video prediction is a pixel-wise dense prediction task to infer future frames based on past frames. Missing appearance details and motion blur are still two major problems for current predictive models, which lead to image distortion and temporal inconsistency. In this paper, we point out the necessity of exploring multi-frequency analysis to deal with the two problems. Inspired by the frequency band decomposition characteristic of Human Vision System (HVS), we propose a video prediction network based on multi-level wavelet analysis to deal with spatial and temporal information in a unified manner. Specifically, the multi-level spatial discrete wavelet transform decomposes each video frame into anisotropic sub-bands with multiple frequencies, helping to enrich structural information and reserve fine details. On the other hand, multi-level temporal discrete wavelet transform which operates on time axis decomposes the frame sequence into sub-band groups of different frequencies to accurately capture multi-frequency motions under a fixed frame rate. Extensive experiments on diverse datasets demonstrate that our model shows significant improvements on fidelity and temporal consistency over state-of-the-art works.

cs.CV↗

arXiv2020

Polarization-Encrypted Orbital Angular Momentum Multiplexed Metasurface Holography

Metasurface holography has the advantage of realizing complex wavefront modulation by thin layers together with the progressive technique of computer-generated holographic imaging. Despite the well-known light parameters, like amplitude, phase, polarization and frequency, the orbital angular momentum (OAM) of a beam can be regarded as another degree of freedom. Here, we propose and demonstrate orbital angular momentum multiplexing at different polarization channels using a birefringent metasurface for holographic encryption. The OAM selective holographic information can only be reconstructed with the exact topological charge and a specific polarization state. By using an incident beam with different topological charges as erasers, we mimic a super-resolution case for the reconstructed image, in analogy to the well-known STED technique in microscopy. The combination of multiple polarization channels together with the orbital angular momentum selectivity provides a higher security level for holographic encryption. Such a technique can be applied for beam shaping, optical camouflage, data storage, and dynamic displays.

physics.optics↗

arXiv2020

EnGN: A High-Throughput and Energy-Efficient Accelerator for Large Graph Neural Networks

Graph neural networks (GNNs) emerge as a powerful approach to process non-euclidean data structures and have been proved powerful in various application domains such as social networks and e-commerce. While such graph data maintained in real-world systems can be extremely large and sparse, thus employing GNNs to deal with them requires substantial computational and memory overhead, which induces considerable energy and resource cost on CPUs and GPUs. In this work, we present a specialized accelerator architecture, EnGN, to enable high-throughput and energy-efficient processing of large-scale GNNs. The proposed EnGN is designed to accelerate the three key stages of GNN propagation, which is abstracted as common computing patterns shared by typical GNNs. To support the key stages simultaneously, we propose the ring-edge-reduce(RER) dataflow that tames the poor locality of sparsely-and-randomly connected vertices, and the RER PE-array to practice RER dataflow. In addition, we utilize a graph tiling strategy to fit large graphs into EnGN and make good use of the hierarchical on-chip buffers through adaptive computation reordering and tile scheduling. Overall, EnGN achieves performance speedup by 1802.9X, 19.75X, and 2.97X and energy efficiency by 1326.35X, 304.43X, and 6.2X on average compared to CPU, GPU, and a state-of-the-art GCN accelerator HyGCN, respectively.

cs.DC↗

arXiv2020

MODMA dataset: a Multi-modal Open Dataset for Mental-disorder Analysis

According to the World Health Organization, the number of mental disorder patients, especially depression patients, has grown rapidly and become a leading contributor to the global burden of disease. However, the present common practice of depression diagnosis is based on interviews and clinical scales carried out by doctors, which is not only labor-consuming but also time-consuming. One important reason is due to the lack of physiological indicators for mental disorders. With the rising of tools such as data mining and artificial intelligence, using physiological data to explore new possible physiological indicators of mental disorder and creating new applications for mental disorder diagnosis has become a new research hot topic. However, good quality physiological data for mental disorder patients are hard to acquire. We present a multi-modal open dataset for mental-disorder analysis. The dataset includes EEG and audio data from clinically depressed patients and matching normal controls. All our patients were carefully diagnosed and selected by professional psychiatrists in hospitals. The EEG dataset includes not only data collected using traditional 128-electrodes mounted elastic cap, but also a novel wearable 3-electrode EEG collector for pervasive applications. The 128-electrodes EEG signals of 53 subjects were recorded as both in resting state and under stimulation; the 3-electrode EEG signals of 55 subjects were recorded in resting state; the audio data of 52 subjects were recorded during interviewing, reading, and picture description. We encourage other researchers in the field to use it for testing their methods of mental-disorder analysis.

cs.DL↗

arXiv2020

A study of resting-state EEG biomarkers for depression recognition

Background: Depression has become a major health burden worldwide, and effective detection depression is a great public-health challenge. This Electroencephalography (EEG)-based research is to explore the effective biomarkers for depression recognition. Methods: Resting state EEG data was collected from 24 major depressive patients (MDD) and 29 normal controls using 128 channel HydroCel Geodesic Sensor Net (HCGSN). To better identify depression, we extracted different types of EEG features including linear features, nonlinear features and functional connectivity features phase lagging index (PLI) to comprehensively analyze the EEG signals in patients with MDD. And using different feature selection methods and classifiers to evaluate the optimal feature sets. Results: Functional connectivity feature PLI is superior to the linear features and nonlinear features. And when combining all the types of features to classify MDD patients, we can obtain the highest classification accuracy 82.31% using ReliefF feature selection method and logistic regression (LR) classifier. Analyzing the distribution of optimal feature set, it was found that intrahemispheric connection edges of PLI were much more than the interhemispheric connection edges, and the intrahemispheric connection edges had a significant differences between two groups. Conclusion: Functional connectivity feature PLI plays an important role in depression recognition. Especially, intrahemispheric connection edges of PLI might be an effective biomarker to identify depression. And statistic results suggested that MDD patients might exist functional dysfunction in left hemisphere.

eess.SP↗

arXiv2020

Dynamic control of mode modulation and spatial multiplexing using hybrid metasurfaces

Designing reconfigurable metasurfaces that can dynamically control scattered electromagnetic waves and work in the near-infrared (NIR) and optical regimes remains a challenging task, which is hindered by the static material property and fixed structures. Phase change materials (PCMs) can provide high contrast optical refractive indexes at high frequencies between amorphous and crystal states, therefore are promising as feasible materials for reconfigurable metasurfaces. Here, we propose a hybrid metasurface that can arbitrarily modulate the complex amplitude of incident light with uniform amplitude and full $2π$ phase coverage by utilizing composite concentric rings (CCRs) with different ratios of gold and PCMs. Our designed metasurface possesses a bi-functionality that is capable of splitting beams or generating vortex beams by thermal switching between metal and semiconductor states of vanadium oxide (VO2), respectively. It can be easily integrated into low loss photonic circuits with an ultra-small footprint. Our metadevice serves as a novel paradigm for active control of beams, which may open new opportunities for signal processing, memory storage, holography, and anti-counterfeiting.

physics.optics↗

arXiv2019

Assessment of Multimodal Choice Behavior for Domestic Passengers Using Bayesian Logistic Regressions: A Case Study in China

This paper investigates the influencing factors in passengers' multimodal traffic choice behaviors and provides a decision-making basis and improvement strategies. By collecting large individual-level data through a comprehensive field survey that was carried out at the major transportation hubs in Xian City in China from the 1st to the 10th of March 2018, we compared 21 variables from the data with four travel modes including air, high-speed rail (HSR), traditional passenger train, and express bus. Among the variables, 12 variables were used as the independent variables after the correlation analysis and the collinearity test, including age, car ownership, ticketing method, travel purpose, travel distance, fare rate, inter-city travel time per hundred kilometers, safety, comfort, punctuality, access time, and departure time. The regression relationships between the travel mode choice and the independent variables were studied using Bayesian binomial logistic regression (BBL) and Bayesian multiple logistic regression (BMNL). The receiver operating characteristic (ROC) curve was applied to analyze the predictive performance of the regression models. The results show that the area under curve (AUC) of the ROC curves for the air travel, high-speed train, passenger train, and express bus using BMNL were 0.9785, 0.9263, 0.8614, and 0.9025, respectively. The AUC value of the ROC curve of the express bus using BBL was 0.8797, which indicated that the passenger travel choice models established by Bayesian logistic regressions had good prediction ability. The application of Bayesian logistic regression in traffic mode choice is beneficial to enriching the assessments of mode-choice behaviors, and the findings of this study can provide a basis for formulating traffic management strategies in practice for shareholders and policymakers.

physics.soc-ph↗

arXiv2019

Simultaneous spectral and spatial modulation for color printing and holography using all-dielectric metasurfaces

Metasurfaces possess the outstanding ability to tailor phase, amplitude and even spectral responses of light with an unprecedented ultrahigh resolution, thus have attracted significant interests. Here, we propose and experimentally demonstrate a novel meta-device that integrates color printing and computer-generated holograms within a single-layer dielectric metasurface by modulating spectral and spatial responses at subwavelength scale, simultaneously. In our design, such metasurface appears as a microscopic color image under white light illumination, while encrypting two different holographic images that can be projected at the far-field when illuminated with red and green laser beams. We choose amorphous silicon dimers and nanofins as building components and use a modified parallel Gerchberg-Saxton algorithm to obtain multiple sub-holograms with arbitrary spatial shapes for image-indexed arrangements. Such a method can further extend the design freedom of metasurfaces. By exploiting spectral and spatial control at the level of individual pixels, multiple sets of independent information can be introduced into a single-layer device, the additional complexity and enlarged information capacity are promising for novel applications such as information security and anti-counterfeiting.

physics.app-ph↗

arXiv2019

HTDet: A Clustering Method using Information Entropy for Hardware Trojan Detection

Hardware Trojans (HTs) have drawn more and more attention in both academia and industry because of its significant potential threat. In this paper, we proposed a novel HT detection method using information entropy based clustering, named HTDet. The key insight of HTDet is that the Trojan usually be inserted in the regions with low controllability and low observability in order to maintain high concealment, which will result in that Trojan logics appear extremely low transitions during the simulation. This means that the logical regions with the low transitions will provide us with much more abundant and more important information for HT detection. Therefore, HTDet applies information theory technology and a typical density-based clustering algorithm called Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to detect all suspicious Trojan logics in circuit under detection (CUD). DBSCAN is an unsupervised learning algorithm, which can improve the applicability of HTDet. Besides, we develop a heuristic test patterns generation method using mutual information to increase the transitions of suspicious Trojan logics. Experimental evaluation with benchmarks demenstrates the effectiveness of HTDet.

cs.IT↗

arXiv2019

Selective Diffraction with Complex Amplitude Modulation by Dielectric Metasurfaces

Metasurfaces have attracted extensive interests due to their ability to locally manipulate optical parameters of light and easy integration to complex optical systems. Particularly, metasurfaces can provide a novel platform for splitting and diffracting light into several beams with desired profile, which is in contrast to traditional gratings. Here, we propose and experimentally demonstrate a novel method for generating independently selective diffraction orders. Our method is based on complex amplitude modulation with ultrathin dielectric metasurfaces. By tailoring the geometric parameters of silicon nanofin structures, we can spatially control the geometric and dynamic phase as well as the amplitude simultaneously. We compare the results with a metasurface that uses a phase-only modulation, to verify such selective diffraction can be solely efficiently achieved with complex amplitude modulation. Besides, the diffraction angles of each order have been measured, which are consistent with standard grating theory. Our developed method for achieving selective diffraction with metasurfaces has potential applications in beam shaping, parallel laser fabrication, and nanoscale optical detection.

physics.optics↗

arXiv2019

Nanoscale polarization manipulation and encryption based on dielectric metasurfaces

Manipulating the polarization of light is highly desired for versatile applications ranging from super resolution, optical trapping, to particle acceleration. The enormous freedom in metasurface design motivates the implementation of polarization control in ultrathin and compact optical systems. However, the majority of proposed strategies based on metasurfaces have been demonstrated only a spatially homogeneous polarization generation, while less attention has been devoted to spatially variant inhomogeneous vector beams. Here, we demonstrate a novel method for generating arbitrary radial and azimuthal polarization beams with high efficiencies of up to 80% by utilizing transmission-type dielectric metasurfaces. Polarization conversion metasurfaces are suitable candidates for the implementation of polarization encryption, which we demonstrate by encoding a hidden image into the spatial polarization distribution. In addition, we show that the image pattern can be modified by appropriate polarization selection of the transmitted light. Such a method may provide a practical technique for a variety of applications such as imaging, encryption and anti-counterfeiting.

physics.optics↗

arXiv2019

Multichannel vectorial holographic display and encryption

Since its invention, holography has emerged as a powerful tool to fully reconstruct the wavefronts of light including all the fundamental properties (amplitude, phase, polarization, wave vector, and frequency). For exploring the full capability for information storage/display and enhancing the encryption security of metasurface holograms, smart multiplexing techniques together with suitable metasurface designs are highly demanded. Here, we integrate multiple polarization manipulation channels for various spatial phase profiles into a single birefringent vectorial hologram by completely avoiding unwanted cross-talk. Multiple independent target phase profiles with quantified phase relations that can process significantly different information in different polarization states are realized within a single metasurface. For our metasurface holograms, we demonstrate high fidelity, large efficiency, broadband operation, and a total of twelve polarization channels. Such multichannel polarization multiplexing can be used for dynamic vectorial holographic display and can provide triple protection for optical security. The concept is appealing for applications of arbitrary spin to angular momentum conversion and various phase modulation/beam shaping elements.

physics.optics↗

arXiv2018

Modified integral imaging reconstruction and encryption using an improved SR reconstruction algorithm

We propose a monospectral image encryption method in which the multispectral color image acquisition by using heterogeneous monospectral cameras. Because the captured monospectral elemental images (EIs) belongs to grayscale image, it is means that the captured EIs can be directly encrypted by the proposed encoding method. Subsequently, the linear cellular automata (CA) and hyperchaotic encoding algorithm are employed to encrypt the captured EIs. Different from previous methods, the proposed method can directly encrypt the multispectral color information rather than having to divide into three color channels (R, G and B), thereby, the proposed method can greatly reduce the encryption calculation.

cs.CR↗

arXiv2015

An optical image encryption scheme based on depth-conversion integral imaging and chaotic maps

Integral imaging-based cryptographic algorithms provides a new way to design secure and robust image encryption schemes. In this paper, we introduce a performance-enhanced image encryption schemes based on depth-conversion integral imaging and chaotic maps, aiming to meet the requirements of secure image transmission. First, the input image is decomposed into an elemental image array (EIA) by utilizing a pinhole array. Then, the obtained image are encrypted by combining the use of cellular automata and chaotic logistic maps. In the image reconstruction process, the conventional computational integral imaging reconstruction (CIIR) technique is a pixel-superposition technique; the resolution of the reconstructed image is dramatically degraded due to the large magnification in the superposition process as the pickup distance increases. The smart mapping technique is introduced to improve the problem of CIIR. A novel property of the proposed scheme is its depth-conversion ability, which converts original elemental images recorded at long distance to ones recorded near the pinhole array and consequently reduce the magnification factor. The results of numerical simulations demonstrate the effectiveness and security of this proposed scheme.

cs.CR↗