PyTorchKR Papers

최신 논문

총 2,153편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2005.06791

High Precision Indoor Navigation for Autonomous Vehicles

Autonomous driving is an important trend of the automotive industry. The continuous research towards this goal requires a precise reference vehicle state estimation under all circumstances in order to develop and test autonomous vehicle functions. However, even when lane-accurate positioning is expected from oncoming technologies, like the L5 GPS band, the question of accurate positioning in roofed areas, e.\,g., tunnels or park houses, still has to be addressed. In this paper, a novel procedure for a reference vehicle state estimation is presented. The procedure includes three main components. First, a robust standstill detection based purely on signals from an Inertial Measurement Unit. Second, a vehicle state estimation by means of statistical filtering. Third, a high accuracy LiDAR-based positioning method that delivers velocity, position and orientation correction data with a mean error of 0.1 m/s, 4.7 cm and 1$^\circ$ respectively. Runtime tests on a CPU indicates the possibility of real-time implementation.

arXiv:2004.00113

Leveraging User-Diversity in Energy-Efficient Edge-Facilitated Collaborative Fog Computing

Motivated by applications such as on-device collaborative neural network inference, this work investigates edge-facilitated collaborative fog computing - in which edge-devices collaborate with each other and with the edge of the network to complete a processing task - to augment the computing capabilities of individual edge-devices while optimizing the collaboration for energy-efficiency. Collaborative computing is modeled using the Map-Reduce distributed computing framework, consisting in two rounds of computations separated by a communication phase. The computing load is optimally distributed among the edge-devices, taking into account their diversity in term of computing and communications capabilities. In addition, edge-devices local parameters such as CPU clock frequency and RF transmit power are also optimized for energy-efficiency. The corresponding optimization problem can be shown to be convex and optimality conditions can be obtained through Lagrange duality theory. A waterfilling-like interpretation for the size of the computing load assigned to each edge-device is given. Numerical experiments demonstrate the benefits of the proposed optimal collaborative-computing scheme over various other schemes in several respects. Most notably, the proposed scheme exhibits increased probability of successfully dealing with heavier computations and/or smaller latency along with energy-efficiency gains of up to two orders of magnitude. Both improvements come from the scheme ability to optimally leverage edge-devices diversity.

arXiv:2003.12726

speckle-tracking: a Software Suite for Ptychographic X-ray Speckle Tracking

In recent years, x-ray speckle tracking techniques have emerged as viable tools for wavefront metrology and sample imaging applications. These methods are based on the measurement of near-field images. Thanks to the simple experimental set-up, high angular sensitivity and compatibility with low coherence sources these methods have been actively developed for use with synchrotron and laboratory light sources. Not only do speckle-tracking techniques give the potential for high resolution imaging, but they also provide rapid and robust characterisation of aberrations of x-ray optical elements, focal spot profiles and the sample position and transmission properties. In order to realise these capabilities, we require software implementations that are equally rapid and robust. To address this need, a software suite has been developed for the "ptychographic x-ray speckle tracking technique" -- an x-ray speckle based method suitable for highly divergent wavefields. The software suite is written in Python 3, with an OpenCL back end for GPU and multi-CPU core processing. It is accessible as a Python module, through the command line or through a graphical user interface and is available as source code under version 3 or later of the GNU General Public License.

arXiv:2002.10708

Controllable Sequence-To-Sequence Neural TTS with LPCNET Backend for Real-time Speech Synthesis on CPU

State-of-the-art sequence-to-sequence acoustic networks, that convert a phonetic sequence to a sequence of spectral features with no explicit prosody prediction, generate speech with close to natural quality, when cascaded with neural vocoders, such as Wavenet. However, the combined system is typically too heavy for real-time speech synthesis on a CPU. In this work we present a sequence-to-sequence acoustic network combined with lightweight LPCNet neural vocoder, designed for real-time speech synthesis on a CPU. In addition, the system allows sentence-level pace and expressivity control at inference time. We demonstrate that the proposed system can synthesize high quality 22 kHz speech in real-time on a general-purpose CPU. In terms of MOS score degradation relative to PCM, the system attained as low as 6.1-6.5% for quality and 6.3- 7.0% for expressiveness, reaching equivalent or better quality when compared to a similar system with a Wavenet vocoder backend.

arXiv:2002.02394

FPGA Acceleration of Sequence Alignment: A Survey

Genomics is changing our understanding of humans, evolution, diseases, and medicines to name but a few. As sequencing technology is developed collecting DNA sequences takes less time thereby generating more genetic data every day. Today the rate of generating genetic data is outpacing the rate of computation power growth. Current sequencing machines can sequence 50 humans genome per day; however, aligning the read sequences against a reference genome and assembling the genome will take 1300 CPU hours. The main step in constructing the genome is aligning the reads against a reference genome. Numerous accelerators have been proposed to accelerate the DNA alignment process. Providing massive parallelism, FPGA-based accelerators have shown great performance in accelerating DNA alignment algorithms. Additionally, FPGA-based accelerators provide better energy efficiency than general-purpose processors. In this survey, we introduce three main DNA alignment algorithms and FPGA-based implementation of these algorithms to accelerate the DNA alignment. We also, compare these three alignment categories and show how accelerators are developing during the time.

arXiv:2001.08259

Energy-Efficient Offloading in Delay-Constrained Massive MIMO Enabled Edge Network Using Data Partitioning

We study a wireless edge-computing system which allows multiple users to simultaneously offload computation-intensive tasks to multiple massive-MIMO access points, each with a collocated multi-access edge computing (MEC) server. Massive-MIMO enables simultaneous uplink transmissions from all users, significantly shortening the data offloading time compared to sequential protocols, and makes the three phases of data offloading, computing, and downloading have comparable durations. Based on this three-phase structure, we formulate a novel problem to minimize a weighted sum of the energy consumption at both the users and the MEC server under a round-trip latency constraint, using a combination of data partitioning, transmit power control and CPU frequency scaling at both the user and server ends. We design a novel nested primal-dual algorithm using two different methods to solve this problem efficiently. Optimized solutions show that for larger requests, more data is offloaded to the MECs to reduce local computation time in order to meet the latency constraint, despite higher energy cost of wireless transmissions. Massive-MIMO channel estimation errors under pilot contamination also causes more data to be offloaded to the MECs. Compared to binary offloading, partial offloading with data partitioning is superior and leads to significant reduction in the overall energy consumption.

arXiv:2001.00316공식 코드 · Papers with Code

Rapid prediction of crucial hotspot interactions for icosahedral viral capsid self-assembly by energy landscape atlasing validated by mutagenesis

Icosahedral viruses have their infectious genome encapsulated by a shell assembled by a multiscale process, starting from an integer multiple of 60 viral capsid or coat protein (VP) monomers. We predict and validate inter-atomic hotspot interactions between VP monomers that are important for the assembly of 3 icosahedral viral capsids: Adeno Associated Virus serotype 2 (AAV2) and Minute Virus of Mice (MVM), both T=1 single stranded DNA viruses, and Bromo Mosaic Virus (BMV), a T=3 single stranded RNA virus. Experimental validation is by in-vitro, site-directed mutagenesis data found in literature. We combine ab-initio predictions at two scales: at the interface-scale, we predict the importance (cruciality) of an interaction for successful subassembly across each interface between VP monomers; and at the capsid-scale, we predict the cruciality of an interface for successful capsid assembly. At the interface-scale, we measure cruciality by changes in the capsid free-energy landscape partition function when an interaction is removed. The partition function computation uses atlases of interface subassembly landscapes, rapidly generated by a novel geometric method and curated opensource software EASAL (efficient atlasing and search of assembly landscapes). At the capsid-scale, cruciality of an interface for successful assembly of the capsid is based on combinatorial entropy. Our study goes from resource-light, multiscale computational predictions of crucial hotspot inter-atomic interactions to validation using data on site-directed mutagenesis' effect on capsid assembly. By reliably and rapidly narrowing down target interactions, (no more than 1.5 hours per interface on a laptop with Intel Core i5-2500K 3.2Ghz CPU and 8GB of RAM) our predictions can inform and reduce time-consuming in-vitro and in-vivo experiments, or more computationally intensive in-silico analyses.

arXiv:1911.10804

An Iterative Interference Cancellation Algorithm for Large Intelligent Surfaces

The Large Intelligent Surface (LIS) concept is a promising technology aiming to revolutionize wireless communication by exploiting spatial multiplexing at its fullest. Despite of its potential, due to the size of the LIS and the large number of antenna elements involved there is a need of decentralized architectures together with distributed algorithms which can reduce the inter-connection data-rate and computational requirement in the Central Processing Unit (CPU). In this article we address the uplink detection problem in the LIS system and propose a decentralize architecture based on panels, which perform local linear processing. We also provide the sum-rate capacity for such architecture and derive an algorithm to obtain the equalizer, which aims to maximize the sum-rate capacity. A performance analysis is also presented, including a comparison to a naive approach based on a reduced form of the matched filter (MF) method. The results shows the superiority of the proposed algorithm.

arXiv:1903.09548

Cost-effective Energy Monitoring of a Zynq-based Real-time System including dual Gigabit Ethernet

The ongoing integration of fine-grained power management features already established in CPU-driven Systems-on-Chip (SoCs) enables both traditional Field Programmable Gate Arrays (FPGAs) and, more recently, hybrid Programmable SoCs (pSoCs) to reach more energy-sensitive application domains (such as, e.g., automotive and robotics). By combining a fixed-function multi-core SoC with flexible, configurable FPGA fabric, the latter can be used to realize heterogeneous Real-time Systems (RTSs) commonly implementing complex application-specific architectures with high computation and communication (I/O) densities. Their dynamic changes in workload, currently active power saving features and thus power consumption require precise voltage and current sensing on all relevant supply rails to enable dependable evaluation of the various power management techniques. In this paper, we propose a low-cost 18-channel 16-bit-resolution measurement (sub-)system capable of 200 kSPS (kilo-samples per second) for instrumentation of current pSoC development boards. To this end, we join simultaneously sampling analog-to-digital converters (ADCs) and analog voltage/current sensing circuitry with a Cortex M7 microcontroller using an SD card for storage. In addition, we propose to include crucial I/O components such as Ethernet PHYs into the power monitoring to gain a holistic view on the RTS's temporal behavior covering not only computation on FPGA and CPUs, but also communication in terms of, e.g., reception of sensor values and transmission of actuation signals. We present an FMC-sized implementation of our measurement system combined with two Gigabit Ethernet PHYs and one HDMI input. Paired with Xilinx' ZC702 development board, we are able to synchronously acquire power traces of a Zynq pSoC and the two PHYs precise enough to identify individual Ethernet frames.

arXiv:1903.04918

A Deep Learning Based Resource Allocation Scheme in Vehicular Communication Systems

In vehicular communications, intracell interference and the stringent latency requirement are challenging issues. In this paper, a joint spectrum reuse and power allocation problem is formulated for hybrid vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. Recognizing the high capacity and low-latency requirements for V2I and V2V links, respectively, we aim to maximize the weighted sum of the capacities and latency requirement. By decomposing the original problem into a classification subproblem and a regression sub-problem, a convolutional neural network (CNN) based approach is developed to obtain real-time decisions on spectrum reuse and power allocation. Numerical results further demonstrate that the proposed CNN can achieve similar performance as the Exhaustive method, while needs only 3.62% of its CPU runtime.

arXiv:1901.01141

BOTDA Fiber Sensor System Based on FPGA Accelerated Support Vector Regression

Brillouin optical time domain analyzer (BOTDA) fiber sensors have shown strong capability in static long haul distributed temperature/strain sensing. However, in applications such as structural health monitoring and leakage detection, real-time measurement is quite necessary. The measurement time of temperature/strain in a BOTDA system includes data acquisition time and post-processing time. In this work, we propose to use hardware accelerated support vector regression (SVR) for the post-processing of the collected BOTDA data. Ideal Lorentzian curves under different temperatures with different linewidths are used to train the SVR model to determine the linear SVR decision function. The performance of SVR is evaluated under different signal-to-noise ratios (SNRs) experimentally. After the model coefficients are determined, algorithm-specific hardware accelerators based on field programmable gate arrays (FPGAs) are used to realize SVR decision function. During the implementation, hardware optimization techniques based on loop dependence analysis and batch processing are proposed to reduce the execution latency. Our FPGA implementations can achieve up to 42x speedup compared with software implementation on an i7-5960x computer. The post-processing time for 96,100 BGSs along 38.44-km FUT is only 0.46 seconds with FPGA board ZCU104, making the post-processing time no longer a limiting factor for dynamic sensing. Moreover, the energy efficiency of our FPGA implementation can reach up to 226.1x higher than software implementation based on CPU.

arXiv:1812.00886

AI Matrix - Synthetic Benchmarks for DNN

Deep neural network (DNN) architectures, such as convolutional neural networks (CNN), involve heavy computation and require hardware, such as CPU, GPU, and AI accelerators, to provide the massive computing power. With the many varieties of AI hardware prevailing on the market, it is often hard to decide which one is the best to use. Thus, benchmarking AI hardware effectively becomes important and is of great help to select and optimize AI hardware. Unfortunately, there are few AI benchmarks available in both academia and industry. Examples are BenchNN[1], DeepBench[2], and Dawn Bench[3], which are usually a collection of typical real DNN applications. While these benchmarks provide performance comparison across different AI hardware, they suffer from a number of drawbacks. First, they cannot adapt to the emerging changes of DNN algorithms and are fixed once selected. Second, they contain tens to hundreds of applications and take very long time to finish running. Third, they are mainly selected from open sources, which are restricted by copyright and are not representable to proprietary applications. In this work, a synthetic benchmarks framework is firstly proposed to address the above drawbacks of AI benchmarks. Instead of pre-selecting a set of open-sourced benchmarks and running all of them, the synthetic approach generates only a one or few benchmarks that best represent a broad range of applications using profiled workload characteristics data of these applications. Thus, it can adapt to emerging changes of new DNN algorithms by re-profiling new applications and updating itself, greatly reduce benchmark count and running time, and strongly represent DNN applications of interests. The generated benchmarks are called AI Matrix, serving as a performance benchmarks matching the statistical workload characteristics of a combination of applications of interests.

arXiv:1811.09366

Prediction of Cytochrome P450-Mediated Metabolism Using a Combination of QSAR Derived Reactivity and Induced Fit Docking

Prediction of metabolism in cytochrome P450s remains to be a crucial yet challenging topic in discovering and designing drugs, agrochemicals and nutritional supplements. The problem is challenging because the rate of P450 metabolism depends upon both the intrinsic chemical reactivity of the site and the protein-ligand geometry that is energetically accessible in the active site of a given P450 isozyme. We have addressed this problem using a two-level screening system. The first level implements an empirical QSAR-based scoring function employing the local chemical motifs to characterize the intrinsic reactivity. The second level uses molecular docking and molecular mechanics to account for the geometrical effects, including induced-fit effects in the protein which can be very important in P450 interactions with ligands. This approach has achieved high accuracy for both the P450 3A4 and 2D6 isoforms. In identifying at least one metabolic site in the top two ranked positions, the prediction rate can reach as high as 92.7% for the test set of isoform 3A4. For the 2D6 isoform, 100% accuracy is achieved on this basic evaluation metric, and, because this active site is considerably smaller and more selective than 3A4, very high precision is attained for full prediction of all metabolic sites. The method also requires considerably less CPU time than our previous efforts, which involved a large number of expensive simulations for each ligand to be evaluated. After screening using the empirical score function, only a few best candidates are left for each ligand, making the number of necessary estimations in the second level very small, which significantly reduces the computation time.

arXiv:1810.11451

5Gperf: signal processing performance for 5G

The 5Gperf project was conducted by Huawei research teams in 2016-17. It was concerned with the acceleration of signal-processing algorithms for a 5G base-station prototype. It improved on already optimized SIMD-parallel CPU algorithms and designed a new software tool for higher programmer productivity when converting MATLAB code to optimized C

arXiv:1809.10170

High Performance Zero-Memory Overhead Direct Convolutions

The computation of convolution layers in deep neural networks typically rely on high performance routines that trade space for time by using additional memory (either for packing purposes or required as part of the algorithm) to improve performance. The problems with such an approach are two-fold. First, these routines incur additional memory overhead which reduces the overall size of the network that can fit on embedded devices with limited memory capacity. Second, these high performance routines were not optimized for performing convolution, which means that the performance obtained is usually less than conventionally expected. In this paper, we demonstrate that direct convolution, when implemented correctly, eliminates all memory overhead, and yields performance that is between 10% to 400% times better than existing high performance implementations of convolution layers on conventional and embedded CPU architectures. We also show that a high performance direct convolution exhibits better scaling performance, i.e. suffers less performance drop, when increasing the number of threads.

arXiv:1806.06171

High Performance Computing in Medical Image Analysis HuSSaR

In our former works we have made serious efforts to improve the performance of medical image analysis methods with using ensemble-based systems. In this paper, we present a novel hardware-based solution for the efficient adoption of our complex, fusion-based approaches for real-time applications. Even though most of the image processing problems and the increasing amount of data have high-performance computing(HPC) demand, there is still a lack of corresponding dedicated HPC solutions for several medical tasks. To widen this bottleneck we have developed a Hybrid Small Size high performance computing Resource (abbreviated by HuSSaR) which efficiently alloys CPU and GPU technologies, mobile and has an own cooling system to support easy mobility and wide applicability. Besides a proper technical description, we include several practical examples from the clinical data processing domain in this work. For more details see also: https://arato.inf.unideb.hu/kovacs.laszlo/research_hybridmicrohpc.html

arXiv:1806.04716

Application of FPGA Acceleration in ADC Performance Calibration

In recent years, high speed and high resolution analog-to-digital converter (ADC) is widely employed in many physical experiments, especially in high precision time and charge measurement. The rapid increasing amount of digitized data demands faster computing. FPGA acceleration has an attracting prospect in data process for its stream process and parallel process feature. In this paper, an ADC performance calibration application based on FPGA acceleration is described. FPGA reads the ADC digitized data stream from PC memory, processes and then writes processed result back to the PC memory. PCIE bus is applied to increase the data transfer speed, and floating point algorithm is applied to improve the accuracy. The test result shows that FPGA acceleration can reduce the processing time of the ADC performance calibration compared with traditional method of C-based CPU processing. This frame of PCIE-based FPGA acceleration method can be applied in analysis and simulation in the future physical experiment for large ADC array, such as CCD camera and waveform digitization readout electronics calibration.

arXiv:1805.04656

An Inner SOCP Approximate Algorithm for Robust Adaptive Beamforming for General-Rank Signal Model

The worst-case robust adaptive beamforming problem for general-rank signal model is considered. Its formulation is to maximize the worst-case signal-to-interference-plus-noise ratio (SINR), incorporating a positive semidefinite constraint on the actual covariance matrix of the desired signal. In the literature, semidefinite program (SDP) techniques, together with others, have been applied to approximately solve this problem. Herein an inner second-order cone program (SOCP) approximate algorithm is proposed to solve it. In particular, a sequence of SOCPs are constructed and solved, while the SOCPs have the nondecreasing optimal values and converge to a locally optimal value (it is in fact a globally optimal value through our extensive simulations). As a result, our algorithm does not use computationally heavy SDP relaxation technique. To validate our inner approximation results, simulation examples are presented, and they demonstrate the improved performance of the new robust beamformer in terms of the averaged cpu-time (indicating how fast the algorithms converge) in a high signal-to-noise region.

arXiv:1706.05935

Speed and biases of Fourier-based pricing choices: A numerical analysis

We compare the CPU effort and pricing biases of seven Fourier-based implementations. Our analyses show that truncation and discretization errors significantly increase as we move away from the Black-Scholes-Merton framework. We rank the speed and accuracy of the competing choices, showing which methods require smaller truncation ranges and which are the most efficient in terms of sampling densities. While all implementations converge well in the Bates jump-diffusion model, Attari's formula is the only Fourier-based method that does not blow up for any Variance Gamma parameter values. In terms of speed, the use of strike vector computations significantly improves the computational burden, rendering both fast Fourier transforms (FFT) and plain delta-probability decompositions inefficient. We conclude that the multi-strike version of the COS method is notably faster than any other implementation, whereas the strike-optimized Carr Madan's formula is simultaneously faster and more accurate than the FFT, thus questioning its use.

arXiv:1703.09129

A Numerical Method for Pricing Discrete Double Barrier Option by Legendre Multiwavelet

In this Article, a fast numerical numerical algorithm for pricing discrete double barrier option is presented. According to Black-Scholes model, the price of option in each monitoring date can be evaluated by a recursive formula upon the heat equation solution. These recursive solutions are approximated by using Legendre multiwavelets as orthonormal basis functions and expressed in operational matrix form. The most important feature of this method is that its CPU time is nearly invariant when monitoring dates increase. Besides, the rate of convergence of presented algorithm was obtained. The numerical results verify the validity and efficiency of the numerical method.