Dongzhu LI Zhijie ZHAN Rei SUMIKAWA Mototsugu HAMADA Atsutake KOSUGE Tadahiro KURODA
A 0.13mJ/prediction with 68.6% accuracy wired-logic deep neural network (DNN) processor is developed in a single 16-nm field-programmable gate array (FPGA) chip. Compared with conventional von-Neumann architecture DNN processors, the energy efficiency is greatly improved by eliminating DRAM/BRAM access. A technical challenge for conventional wired-logic processors is the large amount of hardware resources required for implementing large-scale neural networks. To implement a large-scale convolutional neural network (CNN) into a single FPGA chip, two technologies are introduced: (1) a sparse neural network known as a non-linear neural network (NNN), and (2) a newly developed raster-scan wired-logic architecture. Furthermore, a novel high-level synthesis (HLS) technique for wired-logic processor is proposed. The proposed HLS technique enables the automatic generation of two key components: (1) Verilog-hardware description language (HDL) code for a raster-scan-based wired-logic processor and (2) test bench code for conducting equivalence checking. The automated process significantly mitigates the time and effort required for implementation and debugging. Compared with the state-of-the-art FPGA-based processor, 238 times better energy efficiency is achieved with only a slight decrease in accuracy on the CIFAR-100 task. In addition, 7 times better energy efficiency is achieved compared with the state-of-the-art network-optimized application-specific integrated circuit (ASIC).
Lu ZHAO Qiao-yan WEN Jie ZHANG Zheng-ping JIN
The 2-adic complexity of binary periodic sequences plays an important role in cryptology. In this paper, by means of the usual Fourier transform, we give a simpler form of the upper bound for 2-adic complexity than related result before. For pn-periodic sequences, we discuss the relation between sequences and their Fourier coefficients. Furthermore, based on the relation, we get the lower bound for the number of pn-periodic sequences with given 2-adic complexity.
Xin LIAO Qiaoyan WEN Jie ZHANG
This letter improves two adaptive steganographic methods in Refs. [5], [6], which utilize the remainders of two consecutive pixels to record the information of secret data. Through analysis, we point out that they perform mistakenly under some conditions, and the recipient cannot extract the secret data exactly. We correct these by enlarging the adjusting range of the remainders of two consecutive pixels within the block in the embedding procedure. Furthermore, the readjusting phase in Ref. [6] is improved by allowing every two-pixel block to be fully modified, and then the sender can select the best choice that introduces the smallest embedding distortion. Experimental results show that the improved method not only extracts secret data exactly but also reduces the embedding distortion.
Xiuwen MA Qiaoyan WEN Jie ZHANG Huijuan ZUO
In this letter, by using Whiteman's generalized cyclotomy of order 2 over Zpq, where p, q are twin primes, we construct new perfect Gaussian integer sequences of period pq.
Xin LIAO Qiaoyan WEN Jie ZHANG
In this letter, a novel steganographic method with four-pixel differencing and exploiting modification direction is proposed. Secret data are embedded into each four-pixel block by adaptively applying exploiting modification direction technique. The difference value of the four-pixel block is used to judge whether the pixels in edge areas can tolerate larger changes than those in smooth areas. The readjustment guarantees to extract the secret data exactly and to minimize the embedding distortion. Since the proposed method processes non-overlapping 22 pixels blocks instead of two consecutive pixels, the features of edge can be considered sufficiently. Compared with the previous method, experimental results show that the proposed method provides better performance, i.e., larger embedding capacity and better image quality.
Huijuan ZUO Qiaoyan WEN Xiuwen MA Jie ZHANG
In this letter, we present a general construction of sequence sets with low correlation zone, which is based on finite fields and the balance property of some functions. The construction is more flexible as far as the partition of parameters is concerned. A simple example is also given to interpret the construction.
Pinhui KE Zheng YANG Jie ZHANG
We determine the autocorrelations of the quaternary sequence over F4 and its modified version introduced by Du et al. [X.N. Du et al., Linear complexity of quaternary sequences generated using generalized cyclotomic classes modulo 2p, IEICE Trans. Fundamentals, vol.E94-A, no.5, pp.1214–1217, 2011]. Furthermore, we reveal a drawback in the paper aforementioned and remark that the proof in the paper by Kim et al. can be simplified.
Huijuan WANG Qiaoyan WEN Jie ZHANG
In the construction of a no-linear key-stream generator, self-shrinking is an established way of getting the binary pseudo-random periodic sequences in cryptography design. In this paper, using the theoretical analysis, we mainly study the self-shrinking sequence based on the l-sequence, and the theoretical results reflect its good cryptography properties accurately, such that it has the last period T = pe(p-1)/2 when T is an odd number, and the expected value of its autocorrelation belongs to {0,1/T and the variance is O(T/ln4T). Furthermore, we find that the 2-adic complexity of the self-shrinking sequence based on the l-sequence is large enough to resist the Rational Approximation attack.
Jie ZHANG Chuan XIAO Toyohide WATANABE Yoshiharu ISHIKAWA
Presentation slide composition is an important job for knowledge workers. Instead of starting from scratch, users tend to make new presentation slides by reusing existing ones. A primary challenge in slide reuse is to select desired materials from a collection of existing slides. The state-of-the-art solution utilizes texts and images in slides as well as file names to help users to retrieve the materials they want. However, it only allows users to choose an entire slide as a query but does not support the search for a single element such as a few keywords, a sentence, an image, or a diagram. In this paper, we investigate content-based search for a variety of elements in presentation slides. Users may freely choose a slide element as a query. We propose different query processing methods to deal with various types of queries and improve the search efficiency. A system with a user-friendly interface is designed, based on which experiments are performed to evaluate the effectiveness and the efficiency of the proposed methods.
Zhen GUO Yujie ZHANG Chen SU Jinan XU Hitoshi ISAHARA
Recent work on joint word segmentation, POS (Part Of Speech) tagging, and dependency parsing in Chinese has two key problems: the first is that word segmentation based on character and dependency parsing based on word were not combined well in the transition-based framework, and the second is that the joint model suffers from the insufficiency of annotated corpus. In order to resolve the first problem, we propose to transform the traditional word-based dependency tree into character-based dependency tree by using the internal structure of words and then propose a novel character-level joint model for the three tasks. In order to resolve the second problem, we propose a novel semi-supervised joint model for exploiting n-gram feature and dependency subtree feature from partially-annotated corpus. Experimental results on the Chinese Treebank show that our joint model achieved 98.31%, 94.84% and 81.71% for Chinese word segmentation, POS tagging, and dependency parsing, respectively. Our model outperforms the pipeline model of the three tasks by 0.92%, 1.77% and 3.95%, respectively. Particularly, the F1 value of word segmentation and POS tagging achieved the best result compared with those reported until now.
Lu ZHAO Qiao-yan WEN Jie ZHANG
The linear complexity of quaternary sequences plays an important role in cryptology. In this paper, the minimal polynomial of a class of quaternary sequences with low autocorrelation constructed by generalized cyclotomic sequences pairs is determined, and the linear complexity of the sequences is also obtained.
Xuan ZHANG Qiaoyan WEN Jie ZHANG
In this paper, we propose four new general constructions of LCZ/ZCZ sequence sets based on interleaving technique and affine transformations. A larger family of LCZ/ZCZ sequence sets with longer period are generated by these constructions, which are more flexible among the selection of the alphabet size, the period of the sequences and the length of LCZ/ZCZ, compared with those generated by the known constructions. Especially, two families of the newly constructed sequences can achieve or almost achieve the theoretic bound.
Xuan ZHANG Jing QIN Qiaoyan WEN Jie ZHANG
In this paper, we introduce a construction of 16-QAM sequences based on known binary sequences using multiple sequences, interleaved sequences and Gray mappings. Five kinds of binary sequences of period N are put into the construction to get five kinds of new 16-QAM sequences of period 4N. These resultant sequences have 5-level autocorrelation {0, ±8, ±8N}, where ±8N happens only once each. The distributions of the periodic autocorrelation are also given. These will provide more choices for many applications.
Yu-qian ZHOU Fei GAO Jie ZHANG Qian-yan WEN Zu-ling CHANG
Based on the generalized cyclotomy of order two with respect to n=p1e1+1p2e2+1…ptet+1, where p1, p2, …,pt are pairwise distinct odd primes and e1, e2,…, et are non-negative integers satisfying gcd (piei (pi-1), pjej (pj-1)) = 2 for all i ≠ j, this paper constructs a new family of generalized cyclotomic sequences of order two with length n and investigates their linear complexity. In the view of cascade theory, this paper obtains the linear complexity of a representative sequence.
Huaxi GU Jie ZHANG Zengji LIU Xiaoxing TU
In this letter, a new addressing scheme for hexagonal networks is proposed. Using the new addressing scheme, many routing algorithms designed for networks using square-based topologies such as mesh and torus can also be applied to hexagonal networks. Methods of applying the turn model to hexagonal networks are derived, with some new minimal and partial adaptive routing algorithms obtained. Simulations of the new routing algorithms under different working conditions are carried on hexagonal networks of various sizes. The results show that the proposed algorithms can offer lower packet delay and loss rate than the popular dimension order routing algorithm.
Xiuwen MA Qiaoyan WEN Jie ZHANG Xuan ZHANG
In this paper, we propose new constructions of binary sequences based on an interleaving technique. In our constructions, we make use of any binary sequences with ideal 2-level autocorrelation, a special shift sequence as well as the perfect binary sequence or sequence (0,1,1) in the interleaved structure to get the new sequences. Except for the most autocorrelation values of our new sequences, we find that the unexpected autocorrelation values only occur four or two times in each period no matter how long the period is. We state that the sequences have a good autocorrelation in this case. In particular, the autocorrelation distribution of our sequences is determined.
Xiaojuan ZHU Yang LU Jie ZHANG Zhen WEI
Topological inference is the foundation of network performance analysis and optimization. Due to the difficulty of obtaining prior topology information of wireless sensor networks, we propose routing topology inference, RTI, which reconstructs the routing topology from source nodes to sink based on marking packets and probing locally. RTI is not limited to any specific routing protocol and can adapt to a dynamic and lossy networks. We select topological distance and reconstruction time to evaluate the correctness and effectiveness of RTI and then compare it with PathZip and iPath. Simulation results indicate that RTI maintains adequate reconstruction performance in dynamic and packet loss environments and provides a global routing topology view for wireless sensor networks at a lower reconstruction cost.
Sound source localization is an essential technique in many applications, e.g., speech enhancement, speech capturing and human-robot interaction. However, the performance of traditional methods degrades in noisy or reverberant environments, and it is sensitive to the spatial location of sound source. To solve these problems, we propose a sound source localization framework based on bi-direction interaural matching filter (IMF) and decision weighting fusion. Firstly, bi-directional IMF is put forward to describe the difference between binaural signals in forward and backward directions, respectively. Then, a hybrid interaural matching filter (HIMF), which is obtained by the bi-direction IMF through decision weighting fusion, is used to alleviate the affection of sound locations on sound source localization. Finally, the cosine similarity between the HIMFs computed from the binaural audio and transfer functions is employed to measure the probability of the source location. Constructing the similarity for all the spatial directions as a matrix, we can determine the source location by Maximum A Posteriori (MAP) estimation. Compared with several state-of-the-art methods, experimental results indicate that HIMF is more robust in noisy environments.
Jiao DU Qiaoyan WEN Jie ZHANG Shanqi PANG
In this letter, a property of the characteristic matrix of the Rotation Symmetric Boolean Functions (RSBFs) is characterized, and a sufficient and necessary condition for RSBFs being 1st correlation-immune (1-CI for simplicity) is obtained. This property is applied to construct resilient RSBFs of order 1 (1-resilient for simplicity) on pq variables, where p and q are both prime consistently in this letter. The results show that construction and counting of 1-resilient RSBFs on pq variables are equivalent to solving an equation system and counting the solutions. At last, the counting of all 1-resilient RSBFs on pq variables is also proposed.
Jiao DU Qiaoyan WEN Jie ZHANG Xin LIAO
Orthogonal arrays have important applications in statistics and computer science, as well as in coding theory. In this letter, a new construction method of symmetric orthogonal arrays of strength t is proposed, which is a concatenation of two orthogonal partitions according to a latin square. As far as we know, this is a new construction of symmetric orthogonal arrays of strength t, where t is a given integer. Based on the different latin squares, we also study the enumeration problem of orthogonal partitions, and a lower bound on the count of orthogonal partitions is derived.