Air quality index (AQI) is a non-dimensional index for the description of air quality, and is widely used in air quality management schemes. A novel method for Air Quality Index Forecasting based on Deep Dictionary Learning (AQIF-DDL) and machine vision is proposed in this paper. A sky image is used as the input of the method, and the output is the forecasted AQI value. The deep dictionary learning is employed to automatically extract the sky image features and achieve the AQI forecasting. The idea of learning deeper dictionary levels stemmed from the deep learning is also included to increase the forecasting accuracy and stability. The proposed AQIF-DDL is compared with other deep learning based methods, such as deep belief network, stacked autoencoder and convolutional neural network. The experimental results indicate that the proposed method leads to good performance on AQI forecasting.
Koji KAMMA Yuki ISODA Sarimu INOUE Toshikazu WADA
This paper presents a method for reducing the redundancy in both fully connected layers and convolutional layers of trained neural network models. The proposed method consists of two steps, 1) Neuro-Coding: to encode the behavior of each neuron by a vector composed of its outputs corresponding to actual inputs and 2) Neuro-Unification: to unify the neurons having the similar behavioral vectors. Instead of just pruning one of the similar neurons, the proposed method let the remaining neuron emulate the behavior of the pruned one. Therefore, the proposed method can reduce the number of neurons with small sacrifice of accuracy without retraining. Our method can be applied for compressing convolutional layers as well. In the convolutional layers, the behavior of each channel is encoded by its output feature maps, and channels whose behaviors can be well emulated by other channels are pruned and update the remaining weights. Through several experiments, we comfirmed that the proposed method performs better than the existing methods.
Jing SUN Yi-mu JI Shangdong LIU Fei WU
Software defect prediction (SDP) plays a vital role in allocating testing resources reasonably and ensuring software quality. When there are not enough labeled historical modules, considerable semi-supervised SDP methods have been proposed, and these methods utilize limited labeled modules and abundant unlabeled modules simultaneously. Nevertheless, most of them make use of traditional features rather than the powerful deep feature representations. Besides, the cost of the misclassification of the defective modules is higher than that of defect-free ones, and the number of the defective modules for training is small. Taking the above issues into account, we propose a cost-sensitive and sparse ladder network (CSLN) for SDP. We firstly introduce the semi-supervised ladder network to extract the deep feature representations. Besides, we introduce the cost-sensitive learning to set different misclassification costs for defective-prone and defect-free-prone instances to alleviate the class imbalance problem. A sparse constraint is added on the hidden nodes in ladder network when the number of hidden nodes is large, which enables the model to find robust structures of the data. Extensive experiments on the AEEEM dataset show that the CSLN outperforms several state-of-the-art semi-supervised SDP methods.
Sheng-Hong LIN Jin-Yuan WANG Ying XU Jianxin DAI
This letter investigates the secure transmission improvement scheme for indoor visible light communications (VLC) by using the protected zone. Firstly, the system model is established. For the input signal, the non-negativity and the dimmable average optical intensity constraint are considered. Based on the system model, the secrecy capacity for VLC without considering the protected zone is obtained. After that, the protected zone is determined, and the construction of the protected zone is also provided. Finally, the secrecy capacity for VLC with the protected zone is derived. Numerical results show that the secure performance of VLC improves dramatically by employing the protected zone.
Xin LONG Xiangrong ZENG Chen CHEN Huaxin XIAO Maojun ZHANG
The increase in computation cost and storage of convolutional neural networks (CNNs) severely hinders their applications on limited-resources devices in recent years. As a result, there is impending necessity to accelerate the networks by certain methods. In this paper, we propose a loss-driven method to prune redundant channels of CNNs. It identifies unimportant channels by using Taylor expansion technique regarding to scaling and shifting factors, and prunes those channels by fixed percentile threshold. By doing so, we obtain a compact network with less parameters and FLOPs consumption. In experimental section, we evaluate the proposed method in CIFAR datasets with several popular networks, including VGG-19, DenseNet-40 and ResNet-164, and experimental results demonstrate the proposed method is able to prune over 70% channels and parameters with no performance loss. Moreover, iterative pruning could be used to obtain more compact network.
Chao-Yuan KAO Sangwook PARK Alzahra BADI David K. HAN Hanseok KO
Performance in Automatic Speech Recognition (ASR) degrades dramatically in noisy environments. To alleviate this problem, a variety of deep networks based on convolutional neural networks and recurrent neural networks were proposed by applying L1 or L2 loss. In this Letter, we propose a new orthogonal gradient penalty (OGP) method for Wasserstein Generative Adversarial Networks (WGAN) applied to denoising and despeeching models. WGAN integrates a multi-task autoencoder which estimates not only speech features but also noise features from noisy speech. While achieving 14.1% improvement in Wasserstein distance convergence rate, the proposed OGP enhanced features are tested in ASR and achieve 9.7%, 8.6%, 6.2%, and 4.8% WER improvements over DDAE, MTAE, R-CED(CNN) and RNN models.
Yiling DAI Masatoshi YOSHIKAWA Yasuhito ASANO
The proliferation of Massive Open Online Courses has made it a challenge for the user to select a proper course. We assume a situation in which the user has targeted on the knowledge defined by some knowledge categories. Then, knowing how much of the knowledge in the category is covered by the courses will be helpful in the course selection. In this study, we define a concept of knowledge category coverage and aim to estimate it in a semi-automatic manner. We first model the knowledge category and the course as a set of concepts, and then utilize a taxonomy and the idea of centrality to differentiate the importance of concepts. Finally, we obtain the coverage value by calculating how much of the concepts required in a knowledge category is also taught in a course. Compared with treating the concepts uniformly important, we found that our proposed method can effectively generate closer coverage values to the ground truth assigned by domain experts.
Ning TAI Huan LIN Chao WEI Yongwei LU Chao WANG Kaibo CUI
Since ISAR is widely applied in many occasions and provides high resolution images of the target, ISAR countermeasures are attracting more and more attention. Most of the present methods of deception jamming are not suitable for engineering realization due to the heavy computation load or the large calculation delay. Deception jamming against ISAR requires large computation resource and real-time performance algorithms. Many studies on false target jamming assume that the jammer is able to receive the target echo or transmit the jamming signal to the real target, which is sometimes not possible. How to impose the target property onto the intercepted radar signal is critical to a deception jammer. This paper proposes a jamming algorithm based on parallel convolution and one-bit quantization. The algorithm is able to produce a single false target on ISAR image by the jammer itself. The requirement for computation resource is within the capabilities of current digital signal processors such as FPGA or DSP. The method processes the samples of radar signal in parallel and generates the jamming signal at the rate of ADC data, solving the problem that the real-time performance is not satisfied when the input data rate for convolution is far higher than the clock frequency of FPGA. In order to reduce the computation load of convolution, one-bit quantization is utilized. The complex multiplication is implemented by logical resources, which significantly reduces the consumption of FPGA multipliers. The parallel convolution jamming signal, whose date rate exceeds the FPGA clock rate, is introduced and analyzed in detail. In theory, the bandwidth of jamming signal can be half of the sampling frequency of high speed ADC, making the proposed jamming algorithm able to counter ultra-wideband ISAR signals. The performance and validity of the proposed method are verified by simulations. This jamming method is real-time and capable of producing a false target of large size at the low cost of FPGA device.
Shize KANG Lixin JI Zhenglian LI Xindi HAO Yuehang DING
The goal of cross-lingual entity alignment is to match entities from knowledge graph of different languages that represent the same object in the real world. Knowledge graphs of different languages can share the same ontology which we guess may be useful for entity alignment. To verify this idea, we propose a novel embedding model based on TransC. This model first adopts TransC and parameter sharing model to map all the entities and relations in knowledge graphs to a shared low-dimensional semantic space based on a set of aligned entities. Then, the model iteratively uses reinitialization and soft alignment strategy to perform entity alignment. The experimental results show that, compared with the benchmark algorithms, the proposed model can effectively fuse ontology information and achieve relatively better results.
Takayuki SASAKI Carlos HERNANDEZ GAÑÁN Katsunari YOSHIOKA Michel VAN EETEN Tsutomu MATSUMOTO
Distributed Denial of Service attacks against the application layer (L7 DDoS) are among the most difficult attacks to defend against because they mimic normal user behavior. Some mitigation techniques against L7 DDoS, e.g., IP blacklisting and load balancing using a content delivery network, have been proposed; unfortunately, these are symptomatic treatments rather than fundamental solutions. In this paper, we propose a novel technique to disincentivize attackers from launching a DDoS attack by increasing attack costs. Assuming financially motivated attackers seeking to gain profit via DDoS attacks, their primary goal is to maximize revenue. On the basis of this assumption, we also propose a mitigation solution that requires mining cryptocurrencies to access servers. To perform a DDoS attack, attackers must mine cryptocurrency as a proof-of-work (PoW), and the victims then obtain a solution to the PoW. Thus, relative to attackers, the attack cost increases, and, in terms of victims, the economic damage is compensated by the value of the mined coins. On the basis of this model, we evaluate attacker strategies in a game theory manner and demonstrate that the proposed solution provides only negative economic benefits to attackers. Moreover, we implement a prototype to evaluate performance, and we show that this prototype demonstrates practical performance.
Cuilin CHEN Tsuyoshi SUGIURA Toshihiko YOSHIMASU
This paper presents a 28-GHz-band highly linear stacked-FET power amplifier (PA) IC. A 4-stacked-FET structure is employed for high output power considering the low breakdown voltage of scaled MOSFET transistors. A novel adaptive bias circuit is proposed to dynamically control the gate-to-source bias voltage for amplification MOSFETs. The novel adaptive bias allows the PA to attain high linearity with high back-off efficiency. In addition, the third-order intermodulation distortion (IM3) is improved by a multi-cascode structure. The PA IC is designed, fabricated and fully tested in 56-nm SOI CMOS technology. At a supply voltage of 4 V, the PA IC has achieved an output power of 20.0 dBm with a PAE as high as 38.1% at the 1-dB gain compression point (P1dB). Moreover, PAEs at 3-dB and 6-dB back-off from P1dB are 36.2% and 28.7%, respectively. The PA IC exhibits an output third-order intercept point (OIP3) of 25.0 dBm.
Hikari KOREMURA Haruhiko KANEKO
This paper presents a successive cancellation (SC) decoding of polar codes modified for insertion/deletion/substitution (IDS) error channels, in which insertions and deletions are described by drift values. The recursive calculation of the original SC decoding is modified to include the drift values as stochastic variables. The computational complexity of the modified SC decoding is O (D3) with respect to the maximum drift value D, and O (N log N) with respect to the code length N. The symmetric capacity of polar bit channel is estimated by computer simulations, and frozen bits are determined according to the estimated symmetric capacity. Simulation results show that the decoded error rate of polar code with the modified SC list decoding is lower than that of existing IDS error correction codes, such as marker-based code and spatially-coupled code.
Takuma WAKASA Yoshiki NAGATANI Kenji SAWADA Seiichi SHIN
This paper considers a velocity control problem for merging and splitting maneuvers of vehicle platoons. In this paper, an external device sends velocity commands to some vehicles in the platoon, and the others adjust their velocities autonomously. The former is pinning control, and the latter is consensus control in multi-agent control. We propose a switched pinning control algorithm. Our algorithm consists of three sub-methods. The first is an optimal switching method of pinning agents based on an MLD (Mixed Logical Dynamical) system model and MPC (Model Predictive Control). The second is a representation method for dynamical platoon formation with merging and splitting maneuver. The platoon formation follows the positional relation between vehicles or the formation demand from the external device. The third is a switching reduction method by setting a cost function that penalizes the switching of the pinning agents in the steady-state. Our proposed algorithm enables us to improve the consensus speed. Moreover, our algorithm can regroup the platoons to the arbitrary platoons and control the velocities of the multiple vehicle platoons to each target value.
For embedded systems, verifying both real-time properties and logical validity are important. The embedded system is not only required to the accurate operation but also required to strictly real-time properties. To verify real-time properties is a key problem in model checking. In order to verify real-time properties of assembly program, we develop the simulator to propose the model checking method for verifying assembly programs. Simultaneously, we propose a timed Kripke structure and implement the simulator of the robot's processor to be verified. We propose the timed Kripke structure including the execution time which extends Kripke structure. For the input assembly program, the simulator generates timed Kripke structure by dynamic program analysis. Also, we implement model checker after generating timed Kripke structure in order to verify whether timed Kripke structure satisfies RTCTL formulas. Finally, to evaluate a proposed method, we conduct experiments with the implementation of the verification system. To solve the real problem, we have experimented with real microcontroller software.
Yoshiki TAKAI Mamoru FUKUCHI Chihiro MATSUI Reika KINOSHITA Ken TAKEUCHI
This paper analyzes the optimal SSD configuration including emerging non-volatile memories such as quadruple-level cell (QLC) NAND flash memory [1] and storage class memories (SCMs). First, SSD performance and SSD endurance lifetime of hybrid SSD are evaluated in four configurations: 1) single-level cell (SLC)/QLC NAND flash, 2) SCM/QLC NAND flash, 3) SCM/triple-level cell (TLC)/QLC NAND flash and 4) SCM/TLC NAND flash. Furthermore, these four configurations are compared in limited cost. In case of cold workloads or high total SSD cost assumption, SCM/TLC NAND flash hybrid configuration is recommended in both SSD performance and endurance lifetime. For hot workloads with low total SSD cost assumption, however, SLC/QLC NAND flash hybrid configuration is recommended with emphasis on SSD endurance lifetime. Under the same conditions as above, SCM/TLC/QLC NAND flash tri-hybrid is the best configuration in SSD performance considering cost. In particular, for prxy_0 (write-hot workload), SCM/TLC/QLC NAND flash tri-hybrid achieves 67% higher IOPS/cost than SCM/TLC NAND flash hybrid. Moreover, the configurations with the highest IOPS/cost in each workload and cost limit are picked up and analyzed with various types of SCMs. For all cases except for the case of prxy_1 with high total SSD cost assumption, middle-end SCM (write latency: 1us, read latency: 1us) is recommended in performance considering cost. However, for prxy_1 (read-hot workload) with high total SSD cost assumption, high-end SCM (write latency: 100ns, read latency: 100ns) achieves the best performance.
Han-Ying LIN Chien-Chieh HUANG Wen-Whei CHANG Jen-Tzung CHIEN
This study presents a new method to exploit both accent and grouping structures of music in meter estimation. The system starts by extracting autocorrelation-based features that characterize accent periodicities. Based on the local boundary detection model, we construct grouping features that serve as additional cues for inferring meter. After the feature extraction, a multi-layer cascaded classifier based on neural network is incorporated to derive the most likely meter of input melody. Experiments on 7351 folk melodies in MIDI files indicate that the proposed system achieves an accuracy of 95.76% for classification into nine categories of meters.
Takuya SAKAMOTO Koji NISHIMURA
An analytic expression of the Capon spectrum is derived for two uncorrelated incident signals. On the basis of this theoretical formulation, we discuss the effect of a factor arising from the inner product of mode vectors with respect to the incident angles, which compromises the resolution. We show numerical examples to demonstrate the effect that the inner product of mode vectors has on the shape of the Capon spectrum.
Kaijie ZHOU Huali WANG Peipei CAO Zhangkai LUO
This paper proposes a chirp-BOK modulation scheme for VLF (Very low frequency, 3-30kHz) communication under symmetric alpha-stable (SαS) noise. The atmospheric noise which is the main interference in VLF communication is more accurately characterized as SαS distribution in the previous literatures. Chirp-BOK, one of the chirp spread spectrum (CSS) technologies is widely used for its anti-interference performance and constant envelope properties. However, up-chirp and down-chirp are not strictly orthogonal, the bit error rate (BER) performance of chirp-BOK system is no longer improved with the increase of time-bandwidth product. So in this paper, the influence of non-orthogonal modulation waveform on the system is considered, and the model of the optimal parameters for chirp-BOK is derived from the perspective of minimum BER under gaussian noise and SαS noise respectively. Simulations for chirp-BOK scheme under gaussian noise and SαS noise with different α validate the effectiveness of the proposed method.
Yifan WEI Wanchun LI Yuning GUO Hongshu LIAO
This paper presents a three-dimensional (3D) spatial localization algorithm by using multiple one-dimensional uniform linear arrays (ULA). We first discuss geometric features of the angle-of-arrival (AOA) measurements of the array and present the corresponding principle of spatial cone angle intersection positioning with an angular measurement model. Then, we propose a new positioning method with an analytic study on the geometric dilution of precision (GDOP) of target location in different cases. The results of simulation show that the estimation accuracy of this method can attain the Cramér-Rao Bound (CRB) under low measurement noise.