Fresh Tea Shoot Maturity Estimation (FTSME) is the basement of automatic tea picking technique, determines whether the shoot can be picked. Unfortunately, the ambiguous information among single labels and uncontrollable imaging condition lead to a low FTSME accuracy. A novel Fresh Tea Shoot Maturity Estimating method via multispectral imaging and Deep Label Distribution Learning (FTSME-DLDL) is proposed to overcome these issues. The input is 25-band images, and the output is the corresponding tea shoot maturity label distribution. We utilize the multiple VGG-16 and auto-encoding network to obtain the multispectral features, and learn the label distribution by minimizing the Kullback-Leibler divergence using deep convolutional neural networks. The experimental results show that the proposed method has a better performance on FTSME than the state-of-the-art methods.

System logs record system states and significant events at various critical points to help debug performance issues and failures. Therefore, the rapid and accurate detection of the system log is crucial to the security and stability of the system. In this paper, proposed is a novel attention-based neural network model, which would learn log patterns from normal execution. Concretely, our model adopts a GRU module with attention mechanism to extract the comprehensive and intricate correlations and patterns embedded in a sequence of log entries. Experimental results demonstrate that our proposed approach is effective and achieve better performance than conventional methods.

This paper proposes a dual-polarized metasurface that utilizes multi-layer ceramic capacitors (MLCCs) for radar cross-section (RCS) reduction in the 28GHz band of the quasi-millimeter band. MLCCs are very small in size; therefore, miniaturization of the unit cell structure of the metamaterial can be expected, and the MLCCs can be periodically loaded onto a narrow object. First, the MLCC structure was modeled as a basic structure, and the effective permeability of the MLCC was determined to investigate the influence of the arrangement direction on MLCC interaction. Next, the unit cell structure of the dual-polarized metasurface was designed for an MLCC set on a dielectric substrate. By analyzing the infinite periodic structure and finite structure, the monostatic reduction characteristics, oblique incidence characteristics, and dual-polarization characteristics of the proposed metasurface were evaluated. In the case of the MLCCs arranged in the same direction, the monostatic RCS reduction was approximately 30dB at 29.8GHz, and decreased when the MLCCs were arranged in a checkerboard pattern. The monostatic RCS reductions for the 5 × 5, 10 × 10, and 20 × 20 divisions were roughly the same, i.e., 10.8, 9.9, and 10.3dB, respectively. Additionally, to validate the simulated results, the proposed dual-polarized metasurface was fabricated and measured. The measured results were found to approximately agree with the simulated results, confirming that the RCS can be reduced for dual-polarization operation.

Social media channels, such as Facebook, Twitter, and Instagram, have altered our world forever. People are now increasingly connected than ever and reveal a sort of digital persona. Although social media certainly has several remarkable features, the demerits are undeniable as well. Recent studies have indicated a correlation between high usage of social media sites and increased depression. The present study aims to exploit machine learning techniques for detecting a probable depressed Twitter user based on both, his/her network behavior and tweets. For this purpose, we trained and tested classifiers to distinguish whether a user is depressed or not using features extracted from his/her activities in the network and tweets. The results showed that the more features are used, the higher are the accuracy and F-measure scores in detecting depressed users. This method is a data-driven, predictive approach for early detection of depression or other mental illnesses. This study's main contribution is the exploration part of the features and its impact on detecting the depression level.

Link prediction, the computational problem of determining whether there is a link between two objects, is important in machine learning and data mining. Feature-based link prediction, in which the feature vectors of the two objects are given, is of particular interest because it can also be used for various identification-related problems. Although the factorization machine and the higher-order factorization machine (HOFM) are widely used for feature-based link prediction, they use feature combinations not only across the two objects but also from the same object. Feature combinations from the same object are irrelevant to major link prediction problems such as predicting identity because using them increases computational cost and degrades accuracy. In this paper, we present novel models that use higher-order feature combinations only across the two objects. Since there were no algorithms for efficiently computing higher-order feature combinations only across two objects, we derive one by leveraging reported and newly obtained results of calculating the ANOVA kernel. We present an efficient coordinate descent algorithm for proposed models. We also improve the effectiveness of the existing one for the HOFM. Furthermore, we extend proposed models to a deep neural network. Experimental results demonstrated the effectiveness of our proposed models.

Blockchain is a distributed ledger system composed of a P2P network and is used for a wide range of applications, such as international remittance, inter-individual transactions, and asset conservation. In Blockchain systems, tamper resistance is enhanced by the property of transaction that cannot be changed or deleted by everyone including the creator of the transaction. However, this property also becomes a problem that unintended transaction created by miss operation or secret key theft cannot be corrected later. Due to this problem, once an illegal transaction such as theft occurs, the damage will expand. To suppress the damage, we need countermeasures, such as detecting illegal transaction at high speed and correcting the transaction before approval. However, anomaly detection in the Blockchain at high speed is computationally heavy, because we need to repeat the detection process using various feature quantities and the feature extractions become overhead. In this paper, to accelerate anomaly detection, we propose to cache transaction information necessary for extracting feature in GPU device memory and perform both feature extraction and anomaly detection in the GPU. We also propose a conditional feature extraction method to reduce computation cost of anomaly detection. We employ anomaly detection using K-means algorithm based on the conditional features. When the number of users is one million and the number of transactions is 100 millions, our proposed method achieves 8.6 times faster than CPU processing method and 2.6 times faster than GPU processing method that does not perform feature extraction on the GPU. In addition, the conditional feature extraction method achieves 1.7 times faster than the unconditional method when the number of users satisfying a given condition is 200 thousands out of one million.

This paper describes graph-based Wave Function Collapse algorithm for procedural content generation. The goal of this system is to enable a game designer to procedurally create key content elements in the game level through simple association rule input. To do this, we propose a graph-based data structure that can be easily integrated with a navigation mesh data structure in a three-dimensional world. With our system, if the user inputs the minimum association rule, it is possible to effectively perform procedural content generation in the three-dimensional world. The experimental results show that the Wave Function Collapse algorithm, which is a texture synthesis algorithm, can be extended to non-grid shape content with high controllability and scalability.

In this paper, we propose an interactive system for controlling the pressure while cutting paper with a knife. The purpose is to improve the cutting skill of novices learning the art of paper-cutting. Our system supports skill improvement for novices by measuring and evaluating their cutting pressure in real-time. In this study, we use a knife with a blade attached to a stylus with a pressure sensor, which can measure the pressure, coordinates, and cutting time. We have developed a similar support system using a stylus and a tablet device. This system allows the user to experience the pressure of experts through tracing. Paper-cutting is created by cutting paper with a knife. The practice system in this paper provides practice in an environment more akin to the production of paper cutting. In the first experiment, we observed differences in cutting ability by comparing cutting pressures between novices and experts. As a result, we confirmed that novices cut paper at a higher pressure than experts. We developed a practice system that guides the novices on controlling the pressure by providing information on the cutting pressure values of experts. This system shows the difference in pressure between novices and experts using a synchronous display of color and sound. Using these functions, novices learn to adjust their cutting pressure according to that of experts. Determining the right cutting pressure is a critical skill in the art of paper-cutting, and we aim to improve the same with our system. In the second experiment, we tested the effect of the practice system on the knife device. We compared the changes in cutting pressure with and without our system, the practice methods used in the workshop, and the previously developed stylus-based support system. As a result, we confirmed that practicing with the knife device had a better effect on the novice's skill in controlling cutting pressure than other practice methods.

This paper presents a novel spectrum sensing technique based on selection diversity combining in cognitive radio networks. In general, a selection diversity combining scheme requires a period to select an optimal element, and spectrum sensing requires a period to detect a target signal. We consider that both these periods are required for the spectrum sensing based on selection diversity combining. However, conventional techniques do not consider both the periods. Furthermore, spending a large amount of time in selection and signal detection increases their accuracy. Because the required period for spectrum sensing based on selection diversity combining is the summation of both the periods, their lengths should be considered while developing selection diversity combining based spectrum sensing for a constant period. In reference to this, we discuss the spectrum sensing technique based on selection diversity combining. Numerical examples are shown to validate the effectiveness of the presented design techniques.

This paper presents a novel statistic computation technique for energy detection-based spectrum sensing with multiple antennas. The presented technique computes the statistic for signal detection after combining all the signals. Because the computation of the statistic for all the received signals is not required, the presented technique reduces the computational complexity. Furthermore, the absolute value of all the received signals are combined to prevent the attenuation of the combined signals. Because the statistic computations are not required for all the received signals, the reduction of the computational complexity for signal detection can be expected. Furthermore, the presented technique does not need to choose anything, such as the binary phase rotator in the conventional technique, and therefore, the performance degradation due to wrong choices can be avoided. Numerical examples indicate that the spectrum sensing performances of the presented technique are almost the same as those of conventional techniques despite the complexity of the presented technique being less than that of the conventional techniques.

The choreography realization problem is a design challenge for systems based on service-oriented architecture. In our previous studies, we studied the problem on a case where choreography was given by one or two scenarios and was expressed by an acyclic relation of events; we introduced the notion of re-constructibility as a property of acyclic relations to be satisfied. However, when choreography is defined by multiple scenarios, the resulting behavior cannot be expressed by an acyclic relation. An event structure is composed of an acyclic relation and a conflict relation. Because event structures are a generalization of acyclic relations, a wider class of systems can be expressed by event structures. In this paper, we propose the use of event structures to express choreography, introduce the re-constructibility of event structures, and show a necessary condition for an event structure to be re-constructible.

We propose a deep learning-based model for classifying pathological voices using a convolutional neural network and a feedforward neural network. The model uses combinations of heterogeneous parameters, including mel-frequency cepstral coefficients, linear predictive cepstral coefficients and higher-order statistics. We validate the accuracy of this model using the Massachusetts Eye and Ear Infirmary (MEEI) voice disorder database and the Saarbruecken Voice Database (SVD). Our model achieved an accuracy of 99.3% for MEEI and 75.18% for SVD. This model achieved an accuracy that is 7.18% higher than that of competitive models in previous studies.

In the age of information and communications technology (ICT), not only collecting data but also using such data is provided in various services. It is necessary to ensure data privacy in such services while providing efficient computation and communication complexity. In this paper, we propose the first interval test designed according to the notion of online and offline phases by executing our new empty-set check. Our protocol is proved to ensure both server and client privacy. Furthermore, neither the computational complexity of a client in the online phase nor the communicational complexity from a server to a client depends on the size of the set. As a result, even in a practical situation in which one server receives requests from numerous clients, the waiting time for a client to obtain the result of an interval test can be minimized.

Return-oriented programming (ROP) has been crucial for attackers to evade the security mechanisms of recent operating systems. Although existing ROP detection approaches mainly focus on host-based intrusion detection systems (HIDSes), network-based intrusion detection systems (NIDSes) are also desired to protect various hosts including IoT devices on the network. However, existing approaches are not enough for network-level protection due to two problems: (1) Dynamic approaches take the time with second- or minute-order on average for inspection. For applying to NIDSes, millisecond-order is required to achieve near real time detection. (2) Static approaches generate false positives because they use heuristic patterns. For applying to NIDSes, false positives should be minimized to suppress false alarms. In this paper, we propose a method for statically detecting ROP chains in malicious data by learning the target libraries (i.e., the libraries that are used for ROP gadgets). Our method accelerates its inspection by exhaustively collecting feasible ROP gadgets in the target libraries and learning them separated from the inspection step. In addition, we reduce false positives inevitable for existing static inspection by statically verifying whether a suspicious byte sequence can link properly when they are executed as a ROP chain. Experimental results showed that our method has achieved millisecond-order ROP chain detection with high precision.

Vulnerabilities in hypervisors are crucial in multi-tenant clouds and attractive for attackers because a vulnerability in the hypervisor can undermine all the virtual machine (VM) security. This paper focuses on vulnerabilities in instruction emulators inside hypervisors. Vulnerabilities in instruction emulators are not rare; CVE-2017-2583, CVE-2016-9756, CVE-2015-0239, CVE-2014-3647, to name a few. For backward compatibility with legacy x86 CPUs, conventional hypervisors emulate arbitrary instructions at any time if requested. This design leads to a large attack surface, making it hard to get rid of vulnerabilities in the emulator.This paper proposes FWinst that narrows the attack surface against vulnerabilities in the emulator. The key insight behind FWinst is that the emulator should emulate only a small subset of instructions, depending on the underlying CPU micro-architecture and the hypervisor configuration. FWinst recognizes emulation contexts in which the instruction emulator is invoked, and identifies a legitimate subset of instructions that are allowed to be emulated in the current context. By filtering out illegitimate instructions, FWinst narrows the attack surface. In particular, FWinst is effective on recent x86 micro-architectures because the legitimate subset becomes very small. Our experimental results demonstrate FWinst prevents existing vulnerabilities in the emulator from being exploited on Westmere and Skylake micro-architectures, and the runtime overhead is negligible.

This paper proposes a novel flexible receiver with virtual channels for overloaded multiple-input multiple-output (MIMO) channels. The receiver applies extended rotation matrices proposed in the paper for the flexibility. In addition, adaptive selection of the extended rotation matrices is proposed for further performance improvement. We propose two techniques to reduce the computational complexity of the adaptive selection. As a result, the proposed receiver gives us an option to reduce the complexity with a slight decrease in the transmission performance by changing receiver configuration parameters. A computer simulation reveals that the adaptive selection attains a gain of about 3dB at the BER of 10-3.

We study wireless networked control systems (WNCSs), where controllers (CLs), controlled objects (COs), and other devices are connected through wireless networks. In WNCSs, COs can become unstable due to bursty packet losses and random delays on wireless networks. To reduce these network-induced effects, we utilize the packetized predictive control (PPC) method, where future control vectors to compensate bursty packet losses are generated in the receiving horizon manner, and they are packed into packets and transferred to a CO unit. In this paper, we extend the PPC method so as to compensate random delays as well as bursty packet losses. In the extended PPC method, generating many control vectors improves the robustness against both problems while it increases traffic on wireless networks. Therefore, we consider control vector selection to improve the robustness effectively under the constraint of single packet transmission. We first reconsider the input strategy of control vectors received by COs and propose a control vector selection scheme suitable for the strategy. In our selection scheme, control vectors are selected based on the estimated average and variance of round-trip delays. Moreover, we solve the problem that the CL may misconceive the CO's state due to insufficient information for state estimation. Simulation results show that our selection scheme achieves the higher robustness against both bursty packet losses and delays in terms of the 2-norm of the CO's state.

The static deployment of Virtualized Network Functions (VNFs) introduces 1) significant degradation of Quality of Service (QoS), 2) inefficiency in the network and computing resource utilization, and 3) Network Function Virtualization (NFV)-based services with insufficient scalability, optimality, and flexibility. Caching VNFs is a promising solution to satisfy the dynamic demand to deploy a variety of VNFs and to maximize the performance as well as cost effectiveness. Although the concept of Content Delivery Network (CDN) is popular for efficiently caching and distributing contents, VNF deployment does not realize the benefit of CDN-based caching approaches. The challenges to caching VNFs are 1) to cover the large variety of VNFs and their properties, including the necessity of service chaining, and 2) to achieve high acceptance ratio given the limited availability of resources. This paper proposes Function Delivery Network (FDN), which is a cluster of distributed edge hypervisors for caching VNFs over a Software-Defined Network (SDN). The deployment and quality of the network function can be significantly improved by serving them closer to the end-users from the cached VNFs. FDN introduces a new strategy called Value-based caching that considers 1) the locality of reference and performance parameters of network and edge hypervisors together and 2) a partial deployment of service chains across multiple edge hypervisors for further efficient utilization of hypervisors resources. Evaluations on different patterns of input requests confirm that Value-based caching introduces significant improvement on both QoS and resource utilization in NFV.

Aiming at the problem of low accuracy of target selection in indirect touch input, an indirect multi-touch input device was designed and built. We explored here four indirect touch input techniques which were TarConstant, TarEnlarge, TarAttract, TarEnlargeAttract, and investigated their performance when subjects completing the target selection tasks through comparative experiments. Results showed that TarEnlargeAttract enabled the shortest movement time along with the lowest error rate, 2349.9ms and 10.9% respectively. In terms of learning effect, both TarAttract and TarEnlargeAttract had learning effect on movement time, which indicated that the speed of these two techniques can be improved with training. Finally, the strategy of improving the accuracy of indirect touch input was given, which has reference significance for the interface design of indirect touch input.

This paper addresses the problem of selecting a significant subset of candidate features to use for multiple linear regression. Bertsimas et al. [5] recently proposed the discrete first-order (DFO) algorithm to efficiently find near-optimal solutions to this problem. However, this algorithm is unable to escape from locally optimal solutions. To resolve this, we propose a stochastic discrete first-order (SDFO) algorithm for feature subset selection. In this algorithm, random perturbations are added to a sequence of candidate solutions as a means to escape from locally optimal solutions, which broadens the range of discoverable solutions. Moreover, we derive the optimal step size in the gradient-descent direction to accelerate convergence of the algorithm. We also make effective use of the L2-regularization term to improve the predictive performance of a resultant subset regression model. The simulation results demonstrate that our algorithm substantially outperforms the original DFO algorithm. Our algorithm was superior in predictive performance to lasso and forward stepwise selection as well.