1,720,998 research outputs found

    Low-Power Hardware Accelerator for Sparse Matrix Convolution in Deep Neural Network

    Full text link
    Deep Neural Networks (DNN) have reached an outstanding accuracy in the past years, often going beyond human abilities. Nowadays, DNNs are widely used in many Artificial Intelligence (AI) applications such as computer vision, natural language processing and autonomous driving. However, these incredible performance come at a high computational cost, requiring complex hardware platforms. Therefore, the need for dedicated hardware accelerators able to drastically speed up the execution by preserving a low-power attitude arise. This paper presents innovative techniques able to tackle matrix sparsity in convolutional DNNs due to non-linear activation functions. Developed architectures allow to skip unnecessary operations, like zero multiplications, without sacrificing accuracy or throughput and improving the energy efficiency. Such improvement could enhance the performance of embedded limited-budget battery applications, where cost-effective hardware, accuracy and duration are critical to expanding the deployment of AI

    Approximate-Computing Architectures for Motion Estimation in HEVC

    Full text link
    Video compression deeply relies on motion estima- tion but this crucial step requires power-hungry computational resources. Starting from an existing architecture capable of calculating a high number of Sum of Absolute Differences (SADs), three possible locations for approximate adders are chosen. At each location, different types of approximate adders are implemented and the results are analyzed in terms of error and power saving

    FPGA-based Low-Latency Audio Coprocessor for Networked Music Performance

    Full text link
    Networked Music Performance (NMP) applications are acknowledged to be a particularly challenging field due to their stringent latency requirements and their demand for high audio quality. Most solutions developed in the last decades tried to overcome these obstacles by leveraging software approaches, that can introduce excessive time delays as a consequence of the general-purpose nature of the architectures on which they are implemented. Alternatively, a dedicated audio processor can be employed to minimize the mouth-to-ear latency.This paper presents the ongoing development of an hardware system that exploits an Application-Specific Instruction set Processor (ASIP) implemented on a Field-Programmable Gate Array (FPGA) to accelerate audio sample management. Specifically, a Transport Triggered Architecture (TTA) is being investigated as a processor design that aligns well with the required application domains. Preliminary empirical results indicate that the proposed solution has the potential to achieve extremely low latency, compatible with NMP requirements. Further optimizations and enhancements are actively being pursued to address the yet open challenges posed by NMP applications

    Synthesis, Characterization and Olefin/CO Exchange Reactions of Pyrazolylborato Complexes of Copper(I)

    No full text
    New bis(pyrazolyl)borato olefin complexes of copper(I) of general formula Cu[BH2(3,5-(CF3)2Pz)2]- (olefin) have been prepared (olefins: coe = cyclooctene, van = 4-vinylanisole, clsty = 4-chlorostyrene, tevs = triethylvinylsilane, fn = fumaronitrile). The structures of Cu[BH2(3,5-(CF3)2Pz)2](L), L = coe, van, tevs, fn, have been determined by X-ray diffraction methods. Considering the two N atoms of the bis(pyrazolyl)borate ligand and the midpoint of the C–C double bond of the coordinated olefin, the compounds with L = coe, van and tevs contain a copper atom in a trigonal planar coordination. A coordination polymer with N-coordinated fumaronitrile and tetrahedral coordination of copper atoms is obtained in the case of L = fn. The carbonylation reactions of Cu[BH2(3,5-(CF3)2Pz)2](olefin) (olefin = coe, clsty, van, tevs), Cu[BH2(3,5-(CF3)2Pz)2](olefin) + CO↔Cu[BH2(3,5-(CF3)2Pz)2](CO) + olefin, have been studied gas volumetrically and the thermodynamical parameters of the equilibria for the displacement of the coordinated olefin by carbon monoxide have been determined. These data for copper(I) are compared with those reported in the literature

    Going Beyond Counting First Authors in Author Co-citation Analysis

    Full text link
    The present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
    corecore