Tsuneo KATO Atsushi NAGAI Naoki NODA Jianming WU Seiichi YAMAMOTO
Data-driven untying of a recursive autoencoder (RAE) is proposed for utterance intent classification for spoken dialogue systems. Although an RAE expresses a nonlinear operation on two neighboring child nodes in a parse tree in the application of spoken language understanding (SLU) of spoken dialogue systems, the nonlinear operation is considered to be intrinsically different depending on the types of child nodes. To reduce the gap between the single nonlinear operation of an RAE and intrinsically different operations depending on the node types, a data-driven untying of autoencoders using part-of-speech (PoS) tags at leaf nodes is proposed. When using the proposed method, the experimental results on two corpora: ATIS English data set and Japanese data set of a smartphone-based spoken dialogue system showed improved accuracies compared to when using the tied RAE, as well as a reasonable difference in untying between two languages.
Sukhumarn ARCHASANTISUK Takahiro AOYAGI
Communication reliability and energy efficiency are important issues that have to be carefully considered in WBAN design. Due to the large path loss variation of the WBAN channel, transmission power control, which adaptively adjusts the radio transmit power to suit the channel condition, is considered in this paper. Human motion is one of the dominant factors that affect the channel characteristics in WBAN. Therefore, this paper introduces motion-aware temporal correlation model-based transmission power control that combines human motion classification and transmission power control to provide an effective approach to realizing reliable and energy-efficient WBAN communication. The human motion classification adopted in this study uses only the received signal strength to identify the human motion; no additional tool is required. The knowledge of human motion is then used to accurately estimate the channel condition and suitably select the transmit power. A performance evaluation shows that the proposed method works well both in the low and high WBAN network loads. Compared to using the fixed Tx power of -5dBm, the proposed method had similar packet loss rate but 20-28 and 27-33 percent lower average energy consumption for the low network traffic and high network traffic cases, respectively.
Sae IWATA Kazuaki ISHIKAWA Toshinori TAKAYAMA Masao YANAGISAWA Nozomu TOGAWA
Cell phones with GPS function as well as GPS loggers are widely used and we can easily obtain users' geographic information. Now classifying the measured GPS positions into indoor/outdoor positions is one of the major challenges. In this letter, we propose a robust indoor/outdoor detection method based on sparse GPS measured positions utilizing machine learning. Given a set of clusters of measured positions whose center position shows the user's estimated stayed position, we calculate the feature values composed of: positioning accuracy, spatial features, and temporal feature of measured positions included in every cluster. Then a random forest classifier learns these feature values of the known data set. Finally, we classify the unknown clusters of measured positions into indoor/outdoor clusters using the learned random forest classifier. The experiments demonstrate that our proposed method realizes the maximum F1 measure of 1.000, which classifies measured positions into indoor/outdoor ones with almost no errors.
Sornxayya PHETLASY Satoshi OHZAHATA Celimuge WU Toshihito KATO
Intrusion detection system (IDS) is a device or software to monitor a network system for malicious activity. In terms of detection results, there could be two types of false, namely, the false positive (FP) which incorrectly detects normal traffic as abnormal, and the false negative (FN) which incorrectly judges malicious traffic as normal. To protect the network system, we expect that FN should be minimized as low as possible. However, since there is a trade-off between FP and FN when IDS detects malicious traffic, it is difficult to reduce the both metrics simultaneously. In this paper, we propose a sequential classifiers combination method to reduce the effect of the trade-off. The single classifier suffers a high FN rate in general, therefore additional classifiers are sequentially combined in order to detect more positives (reduce more FN). Since each classifier can reduce FN and does not generate much FP in our approach, we can achieve a reduction of FN at the final output. In evaluations, we use NSL-KDD dataset, which is an updated version of KDD Cup'99 dataset. WEKA is utilized as a classification tool in experiment, and the results show that the proposed approach can reduce FN while improving the sensitivity and accuracy.
Abu Hena Al MUKTADIR Takaya MIYAZAWA Pedro MARTINEZ-JULIA Hiroaki HARAI Ved P. KAFLE
In this paper, we propose a method for automatic virtual resource allocation by using a multi-target classification-based scheme (MTCAS). In our method, an Infrastructure Provider (InP) bundles its CPU, memory, storage, and bandwidth resources as Network Elements (NEs) and categorizes them into several types in accordance to their function, capabilities, location, energy consumption, price, etc. MTCAS is used by the InP to optimally allocate a set of NEs to a Virtual Network Operator (VNO). Such NEs will be subject to some constraints, such as the avoidance of resource over-allocation and the satisfaction of multiple Quality of Service (QoS) metrics. In order to achieve a comparable or higher prediction accuracy by using less training time than the available ensemble-based multi-target classification (MTC) algorithms, we propose a majority-voting based ensemble algorithm (MVEN) for MTCAS. We numerically evaluate the performance of MTCAS by using the MVEN and available MTC algorithms with synthetic training datasets. The results indicate that the MVEN algorithm requires 70% less training time but achieves the same accuracy as the related ensemble based MTC algorithms. The results also demonstrate that increasing the amount of training data increases the efficacy ofMTCAS, thus reducing CPU and memory allocation by about 33% and 51%, respectively.
Seong Jin CHOE Ju Sang LEE Sung Sik PARK Sang Dae YU
This paper presents an ultra-low-power class-AB bulk-driven operational transconductance amplifier operating in the subthreshold region. Employing the partial positive feedback in current mirrors, the effective transconductance and output voltage swing are enhanced considerably without additional power consumption and layout area. Both traditional and proposed OTAs are designed and simulated for a 180 nm CMOS process. They dissipate an ultra low power of 192 nW. The proposed OTA features not only a DC gain enhancement of 14 dB but also a slew rate improvement of 200%. In addition, the improved gain leads to a 5.3 times wider unity-gain bandwidth than that of the traditional OTA.
Eeva-Sofia HAUKIPURO Ville KOLEHMAINEN Janne MYLLÄRINEN Sebastian REMANDER Janne SALO Tuomas TAKKO Le Ngu NGUYEN Stephan SIGG Rainhard Dieter FINDLING
Biometric authentication, namely using biometric features for authentication is gaining popularity in recent years as further modalities, such as fingerprint, iris, face, voice, gait, and others are exploited. We explore the effectiveness of three simple Electroencephalography (EEG) related biometric authentication tasks, namely resting, thinking about a picture, and moving a single finger. We present details of the data processing steps we exploit for authentication, including extracting features from the frequency power spectrum and MFCC, and training a multilayer perceptron classifier for authentication. For evaluation purposes, we record an EEG dataset of 27 test subjects. We use three setups, baseline, task-agnostic, and task-specific, to investigate whether person-specific features can be detected across different tasks for authentication. We further evaluate, whether different tasks can be distinguished. Our results suggest that tasks are distinguishable, as well as that our authentication approach can work both exploiting features from a specific, fixed, task as well as using features across different tasks.
As the data size of Web-related multi-label classification problems continues to increase, the label space has also grown extremely large. For example, the number of labels appearing in Web page tagging and E-commerce recommendation tasks reaches hundreds of thousands or even millions. In this paper, we propose a graph partitioning tree (GPT), which is a novel approach for extreme multi-label learning. At an internal node of the tree, the GPT learns a linear separator to partition a feature space, considering approximate k-nearest neighbor graph of the label vectors. We also developed a simple sequential optimization procedure for learning the linear binary classifiers. Extensive experiments on large-scale real-world data sets showed that our method achieves better prediction accuracy than state-of-the-art tree-based methods, while maintaining fast prediction.
Lina GONG Shujuan JIANG Qiao YU Li JIANG
Heterogeneous defect prediction (HDP) is to detect the largest number of defective software modules in one project by using historical data collected from other projects with different metrics. However, these data can not be directly used because of different metrics set among projects. Meanwhile, software data have more non-defective instances than defective instances which may cause a significant bias towards defective instances. To completely solve these two restrictions, we propose unsupervised deep domain adaptation approach to build a HDP model. Specifically, we firstly map the data of source and target projects into a unified metric representation (UMR). Then, we design a simple neural network (SNN) model to deal with the heterogeneous and class-imbalanced problems in software defect prediction (SDP). In particular, our model introduces the Maximum Mean Discrepancy (MMD) as the distance between the source and target data to reduce the distribution mismatch, and use the cross-entropy loss function as the classification loss. Extensive experiments on 18 public projects from four datasets indicate that the proposed approach can build an effective prediction model for heterogeneous defect prediction (HDP) and outperforms the related competing approaches.
Song BIAN Masayuki HIROMOTO Takashi SATO
In this work, we provide the first practical secure email filtering scheme based on homomorphic encryption. Specifically, we construct a secure naïve Bayesian filter (SNBF) using the Paillier scheme, a partially homomorphic encryption (PHE) scheme. We first show that SNBF can be implemented with only the additive homomorphism, thus eliminating the need to employ expensive fully homomorphic schemes. In addition, the design space for specialized hardware architecture realizing SNBF is explored. We utilize a recursive Karatsuba Montgomery structure to accelerate the homomorphic operations, where multiplication of 2048-bit integers are carried out. Through the experiment, both software and hardware versions of the SNBF are implemented. On software, 104-105x runtime and 103x storage reduction are achieved by SNBF, when compared to existing fully homomorphic approaches. By instantiating the designed hardware for SNBF, a further 33x runtime and 1919x power reduction are achieved. The proposed hardware implementation classifies an average-length email in under 0.5s, which is much more practical than existing solutions.
Yusuke AIKAWA Koji NUIDA Masaaki SHIRASE
In 2017, Shirase proposed a variant of Elliptic Curve Method combined with Complex Multiplication method for generating certain special kinds of elliptic curves. His algorithm can efficiently factorize a given composite integer when it has a prime factor p of the form 4p=1+Dv2 for some integer v, where -D is an auxiliary input integer called a discriminant. However, there is a disadvantage that the previous method works only for restricted cases where the class polynomial associated to -D has degree at most two. In this paper, we propose a generalization of the previous algorithm to the cases of class polynomials having arbitrary degrees, which enlarges the class of composite integers factorizable by our algorithm. We also extend the algorithm to more various cases where we have 4p=t2+Dv2 and p+1-t is a smooth integer.
Extreme multi-label classification methods have been widely used in Web-scale classification tasks such as Web page tagging and product recommendation. In this paper, we present a novel graph embedding method called “AnnexML”. At the training step, AnnexML constructs a k-nearest neighbor graph of label vectors and attempts to reproduce the graph structure in the embedding space. The prediction is efficiently performed by using an approximate nearest neighbor search method that efficiently explores the learned k-nearest neighbor graph in the embedding space. We conducted evaluations on several large-scale real-world data sets and compared our method with recent state-of-the-art methods. Experimental results show that our AnnexML can significantly improve prediction accuracy, especially on data sets that have a larger label space. In addition, AnnexML improves the trade-off between prediction time and accuracy. At the same level of accuracy, the prediction time of AnnexML was up to 58 times faster than that of SLEEC, a state-of-the-art embedding-based method.
Mayu OTANI Atsushi NISHIDA Yuta NAKASHIMA Tomokazu SATO Naokazu YOKOYA
Finding important regions is essential for applications, such as content-aware video compression and video retargeting to automatically crop a region in a video for small screens. Since people are one of main subjects when taking a video, some methods for finding important regions use a visual attention model based on face/pedestrian detection to incorporate the knowledge that people are important. However, such methods usually do not distinguish important people from passers-by and bystanders, which results in false positives. In this paper, we propose a deep neural network (DNN)-based method, which classifies a person into important or unimportant, given a video containing multiple people in a single frame and captured with a hand-held camera. Intuitively, important/unimportant labels are highly correlated given that corresponding people's spatial motions are similar. Based on this assumption, we propose to boost the performance of our important/unimportant classification by using conditional random fields (CRFs) built upon the DNN, which can be trained in an end-to-end manner. Our experimental results show that our method successfully classifies important people and the use of a DNN with CRFs improves the accuracy.
The research on inertial sensor based human action detection and recognition (HADR) is a new area in machine learning. We propose a novel time sequence based interval convolutional neutral networks framework for HADR by combining interesting interval proposals generator and interval-based classifier. Experiments demonstrate the good performance of our method.
In this letter, we propose a static wear leveling technique, called Recency-based Wear Leveling (RbWL). The basic idea of RbWL is to execute static wear leveling at minimum levels, because the frequent migrations of cold data by static wear leveling cause significant overhead in a NAND flash memory system. RbWL adjusts the execution frequency according to a threshold value that reflects the lifetime difference of the hot/cold blocks and the total lifetime of the NAND flash memory system. The evaluation results show that RbWL improves the lifetime of NAND flash memory systems by 52%, and it also reduces the overhead of wear leveling from 8% to 42% and from 13% to 51%, in terms of the number of erase operations and the number of page migrations of valid pages, respectively, compared with other algorithms.
Yuehua WANG Zhinong ZHONG Anran YANG Ning JING
Review rating prediction is an important problem in machine learning and data mining areas and has attracted much attention in recent years. Most existing methods for review rating prediction on Location-Based Social Networks only capture the semantics of texts, but ignore user information (social links, geolocations, etc.), which makes them less personalized and brings down the prediction accuracy. For example, a user's visit to a venue may be influenced by their friends' suggestions or the travel distance to the venue. To address this problem, we develop a review rating prediction framework named TSG by utilizing users' review Text, Social links and the Geolocation information with machine learning techniques. Experimental results demonstrate the effectiveness of the framework.
MathML is a standard markup language for describing math expressions. MathML consists of two sets of elements: Presentation Markup and Content Markup. The former is widely used to display math expressions in Web pages, while the latter is more suited to the calculation of math expressions. In this letter, we focus on the former and consider classifying Presentation MathML expressions. Identifying the classes of given Presentation MathML expressions is helpful for several applications, e.g., Presentation to Content MathML conversion, text-to-speech, and so on. We propose a method for classifying Presentation MathML expressions by using multilayer perceptron. Experimental results show that our method classifies MathML expressions with high accuracy.
Yizhou JIANG Sai HUANG Yixin ZHANG Zhiyong FENG Di ZHANG Celimuge WU
This letter proposes a novel modulation classification method for overlapped sources named LRGP involving multinomial logistic regression (MLR) and multi-gene genetic programming (MGGP). MGGP based feature engineering is conducted to transform the cumulants of the received signals into highly discriminative features and a MLR based classifier is trained to identify the combination of the modulation formats of the overlapped sources instead of signal separation. Extensive simulations demonstrate that LRGP yields superior performance compared with existing methods.
Yuto KUROSAKI Masayoshi OHTA Hidetaka ITO Hiroomi HIKAWA
This paper discusses the effect of pre-grouping on vector classification based on the self-organizing map (SOM). The SOM is an unsupervised learning neural network, and is used to form clusters of vectors using its topology preserving nature. The use of SOMs for practical applications, however, may pose difficulties in achieving high recognition accuracy. For example, in image recognition, the accuracy is degraded due to the variation of lighting conditions. This paper considers the effect of pre-grouping of feature vectors on such types of applications. The proposed pre-grouping functionality is also based on the SOM and introduced into a new parallel configuration of the previously proposed SOM-Hebb classifers. The overall system is implemented and applied to position identification from images obtained in indoor and outdoor settings. The system first performs the grouping of images according to the rough representation of the brightness profile of images, and then assigns each SOM-Hebb classifier in the parallel configuration to one of the groups. Recognition parameters of each classifier are tuned for the vectors belonging to its group. Comparison between the recognition systems with and without the grouping shows that the grouping can improve recognition accuracy.
We present a new action classification method for skeletal sequence data. The proposed method is based on simple nonparametric feature matching without a learning process. We first augment the training dataset to implicitly construct an exponentially increasing number of training sequences, which can be used to improve the generalization power of the proposed action classifier. These augmented training sequences are matched to the test sequence with the relaxed dynamic time warping (DTW) technique. Our relaxed formulation allows the proposed method to work faster and with higher efficiency than the conventional DTW-based method using a non-augmented dataset. Experimental results show that the proposed approach produces effective action classification results for various scales of real datasets.