Face recognition across pose is generally tackled by either 2D based or 3D based approaches. The 2D-based often require a training set from which the cross-pose multi-view relationship can be learned and applied for recognition. The 3D based are mostly composed of 3D surface reconstruction of each gallery face, synthesis of 2D images of novel views using the reconstructed model, and match of the synthesized images to the probes. The depth information provides crucial information for arbitrary poses but more methods are yet to be developed. Extended from a latest face reconstruction method using a single 3D reference model and a frontal registered face, this study focuses on using the reconstructed 3D face for recognition. The recognition performance varies with poses, the closer to the front, the better. Several ways to improve the performance are attempted, including different numbers of fiducial points for alignment, multiple reference models considered in the reconstruction phase, and both frontal and profile poses available in the gallery. These attempts make this approach competitive to the state-of-the-art methods.

A class-E power amplifier (PA) with novel dynamic biasing scheme is proposed to enhance power added efficiency (PAE) over a wide power range. A look-up table (LUT) adjusts input power and drain supply voltage simultaneously to keep switch mode condition of a power transistor and to optimize the PAE. Experimental results show that the class-E PA using the proposed scheme with harmonic suppression filter gives the PAE higher than 80% over 8.5,dB range with less than 40,dBc harmonic suppression.

A Balanced Neighborhood Classifier (BNEC) is proposed for class imbalanced data. This method is not only well positioned to capture the class distribution information, but also has the good merits of high-fitting-performance and simplicity. Experiments on both synthetic and real data sets show its effectiveness.

In this paper, sequential prediction is studied. The typical assumptions about the probabilistic model in sequential prediction are following two cases. One is the case that a certain probabilistic model is given and the parameters are unknown. The other is the case that not a certain probabilistic model but a class of probabilistic models is given and the parameters are unknown. If there exist some parameters and some models such that the distributions that are identified by them equal the source distribution, an assumed model or a class of models can represent the source distribution. This case is called that specifiable condition is satisfied. In this study, the decision based on the Bayesian principle is made for a class of probabilistic models (not for a certain probabilistic model). The case that specifiable condition is not satisfied is studied. Then, the asymptotic behaviors of the cumulative logarithmic loss for individual sequence in the sense of almost sure convergence and the expected loss, i.e. redundancy are analyzed and the constant terms of the asymptotic equations are identified.

One of the major applications of contrast-enhanced ultrasound (CEUS) is lesion classification. After contrast agents are administered, it is possible to identify a lesion type from its enhancement pattern. However, CEUS image reading is not easy because there are various types of enhancement patterns even for the same type of lesion, and clear classification criteria have not yet been defined. Some studies have used conventional time intensity curves (TICs), which show the vessel dynamics of a lesion. It is possible to predict lesion type from the TIC parameters, such as the coefficients obtained by curve fitting, peak intensity, flow rate and time to peak. However, these parameters are not always provide sufficient accuracy. In this paper, we prepare 1D Haar-like features which describe intensity changes in a TIC and adopt the Adaboost machine learning technique, which eases understanding of which features are useful. Hyperparameters of weak classifiers, e.g., the step size of a Haar-like filter length and threshold for output of the filter, are optimized by searching for those parameters that give the best accuracy. We evaluate the proposed method using 36 focal splenic lesions in canines 16 of which were benign and 20 malignant. The accuracies were 91.7% (33/36) when inspected by an experienced veterinarian, 75.0% (27/36) by linear discriminant analysis (LDA) using conventional three TIC parameters: time to peak, area under curve and peak intensity, and 91.7% (33/36) using our proposed method. McNemar testing shows the p-value to be less than 0.05 between the proposed method and LDA. This result shows the statistical significance of differences between the proposed method and the conventional TIC analysis method using LDA.

This paper describes the design method of a broadband CMOS stacked power amplifier using harmonic control over wide bandwidths in a 0.11,$mu $m standard CMOS process. The high-efficiency can be obtained over wide bandwidths by designing a load impedance circuit as purely reactive as possible to the harmonics with broadband fundamental matching, which is based on continuous Class-F mode of operation. Furthermore, the stacked topology overcomes the low breakdown voltage limit of CMOS transistor and increases output impedance. With a 5-V supply and a fixed matching circuitry, the suggested power amplifier (PA) achieves a saturated output power of over 26.7,dBm and a drain efficiency of over 38% from 1.6,GHz to 2.2,GHz. In W-CDMA modulation signal measurements, the PA generates linear power and power added efficiency of over 23.5,dBm and 33% (@ACLR $=-33$,dBc).

In this paper, we propose a texture descriptor based on amplitude distribution and phase distribution of the discrete Fourier transform (DFT) of an image. One dimensional DFT is applied to all the rows and columns of an image. Histograms of the amplitudes and gradients of the phases between adjacent rows/columns are computed as the feature descriptor, which is called aggregated DFT (ADFT). ADFT can be easily combined with completed local binary pattern (CLBP). The combined feature captures both global and local information of the texture. ADFT is designed for isotropic textures and demonstrated to be effective for roughness classification of castings. Experimental results show that the amplitude part of ADFT is also discriminative in describing anisotropic textures and it can be used as a complementary descriptor of local texture descriptors such as CLBP.

Internet traffic classification is an essential step for stable service provision. The payload signature classifier is considered a reliable method for Internet traffic classification but is prohibitively computationally expensive for real-time handling of large amounts of traffic on high-speed networks. In this paper, we describe several design techniques to minimize the search space of traffic classification and improve the processing speed of the payload signature classifier. Our suggestions are (1) selective matching algorithms based on signature type, (2) signature reorganization using hierarchical structure and traffic locality, and (3) early packet sampling in flow. Each can be applied individually, or in any combination in sequence. The feasibility of our selections is proved via experimental evaluation on traffic traces of our campus and a commercial ISP. We observe 2 to 5 times improvement in processing speed against the untuned classification system and Snort Engine, while maintaining the same level of accuracy.

A Decision Diagram Machine (DDM) is a special-purpose processor that has special instructions to evaluate a decision diagram. Since the DDM uses only a limited number of instructions, it is faster than the general-purpose Micro Processor Unit (MPU). Also, the architecture for the DDM is much simpler than that for an MPU. This paper presents a packet classifier using a parallel EVMDD (k) machine. To reduce computation time and code size, first, a set of rules for a packet classifier is partitioned into groups. Then, the parallel EVMDD (k) machine evaluates them. To further speed-up for the standard EVMDD (k) machine, we propose the prefetching EVMDD (k) machine which reads both the index and the jump address at the same time. The prefetching EVMDD (k) machine is 2.4 times faster than the standard one using the same memory size. We implemented a parallel prefetching EVMDD (k) machine consisting of 30 machines on an FPGA, and compared it with the Intel's Core i5 microprocessor running at 1.7GHz. Our parallel machine is 15.1-77.5 times faster than the Core i5, and it requires only 8.1-58.5 percents of the memory for the Core i5.

Detecting a variety of anomalies caused by attacks or accidents in computer networks has been one of the real challenges for both researchers and network operators. An effective technique that could quickly and accurately detect a wide range of anomalies would be able to prevent serious consequences for system security or reliability. In this article, we characterize detection techniques on the basis of learning models and propose an unsupervised learning model for real-time anomaly detection in computer networks. We also conducted a series of experiments to examine capabilities of the proposed model by employing three well-known machine learning algorithms, namely multivariate normal distribution, k-nearest neighbor, and one-class support vector machine. The results of these experiments on real network traffic suggest that the proposed model is a promising solution and has a number of flexible capabilities to detect several types of anomalies in real time.

The dynamic characteristics of the class E power amplifier with frequency modulation are derived. Such an analysis is essential for designing amplitude and frequency modulated amplifier systems such as an EER scheme. Conventionally, an analytical expression for the frequency response of a frequency modulated class E amplifier has not been derived yet. This omission is rectified here by modeling the circuit with both a low-frequency model and a high-frequency model. Further, a time domain waveform is derived from the frequency domain transfer function for some typical time varying drive signals. The analytical results for the frequency response of a 1-MHz class E amplifier are shown to match PSpice simulations and measured values well.

Software products are increasingly complex, so it is becoming more difficult to find and correct bugs in large programs. Software developers rely on bug reports to fix bugs; thus, bug-tracking tools have been introduced to allow developers to upload, manage, and comment on bug reports to guide corrective software maintenance. However, the very high frequency of duplicate bug reports means that the triagers who help software developers in eliminating bugs must allocate large amounts of time and effort to the identification and analysis of these bug reports. In addition, classifying bug reports can help triagers arrange bugs in categories for the fixers who have more experience for resolving historical bugs in the same category. Unfortunately, due to a large number of submitted bug reports every day, the manual classification for these bug reports increases the triagers' workload. To resolve these problems, in this study, we develop a novel technique for automatic duplicate detection and classification of bug reports, which reduces the time and effort consumed by triagers for bug fixing. Our novel technique uses a support vector machine to check whether a new bug report is a duplicate. The concept profile is also used to classify the bug reports into related categories in a taxonomic tree. Finally, we conduct experiments that demonstrate the feasibility of our proposed approach using bug reports extracted from the large-scale open source project Mozilla.

We address the problem of estimating the difference between two probability densities. A naive approach is a two-step procedure that first estimates two densities separately and then computes their difference. However, such a two-step procedure does not necessarily work well because the first step is performed without regard to the second step and thus a small error in the first stage can cause a big error in the second stage. Recently, a single-shot method called the least-squares density-difference (LSDD) estimator has been proposed. LSDD directly estimates the density difference without separately estimating two densities, and it was demonstrated to outperform the two-step approach. In this paper, we propose a variation of LSDD called the constrained least-squares density-difference (CLSDD) estimator, and theoretically prove that CLSDD improves the accuracy of density difference estimation for correctly specified parametric models. The usefulness of the proposed method is also demonstrated experimentally.

Classification tasks in computer vision and brain-computer interface research have presented several applications such as biometrics and cognitive training. However, like in any other discipline, determining suitable representation of data has been challenging, and recent approaches have deviated from the familiar form of one vector for each data sample. This paper considers a kernel between vector sets, the mean polynomial kernel, motivated by recent studies where data are approximated by linear subspaces, in particular, methods that were formulated on Grassmann manifolds. This kernel takes a more general approach given that it can also support input data that can be modeled as a vector sequence, and not necessarily requiring it to be a linear subspace. We discuss how the kernel can be associated with the Projection kernel, a Grassmann kernel. Experimental results using face image sequences and physiological signal data show that the mean polynomial kernel surpasses existing subspace-based methods on Grassmann manifolds in terms of predictive performance and efficiency.

A multi-source Tri-Training transfer learning algorithm is proposed by integrating transfer learning and semi-supervised learning. First, multiple weak classifiers are respectively trained by using both weighted source and target training samples. Then, based on the idea of co-training, each target testing sample is labeled by using trained weak classifiers and the sample with the same label is selected as the high-confidence sample to be added into the target training sample set. Finally, we can obtain a target domain classifier based on the updated target training samples. The above steps are iterated till the high-confidence samples selected at two successive iterations become the same. At each iteration, source training samples are tested by using the target domain classifier and the samples tested as correct continue with training, while the weights of samples tested as incorrect are lowered. Experimental results on text classification dataset have proven the effectiveness and superiority of the proposed algorithm.

This paper proposes an ultra-low-power 5.5-GHz PLL which employs the new divide-by-4 injection-locked frequency divider (ILFD) and a class-C VCO with linearity-compensated varactor for low supply voltage operation. A forward-body-biasing (FBB) technique can decrease threshold voltage of MOS transistors, which can improve operation frequency and can widen the lock range of the ILFD. The FBB is also employed for linear-frequency-tuning of VCO under low supply voltage of 0.5V. The double-switch injection technique is also proposed to widen the lock range of the ILFD. The digital calibration circuit is introduced to control the lock-range of ILFD automatically. The proposed PLL was fabricated in a 65nm CMOS process. With a 34.3-MHz reference, it shows a 1-MHz-offset phase noise of -106dBc/Hz at 5.5GHz output. The supply voltage is 0.54V for divider and 0.5V for other components. Total power consumption is 0.95mW.

We propose a novel motion segmentation method based on a Clausius Normalized Field (CNF), a probabilistic model for treating time-varying imagery, which estimates entropy variations by observing the entropy definitions of Clausius and Boltzmann. As pixels of an image are viewed as a state of lattice-like molecules in a thermodynamic system, estimating entropy variations of pixels is the same as estimating their degrees of disorder. A greater increase in entropy means that a pixel has a higher chance of belonging to moving objects rather than to the background, because of its higher disorder. In addition to these homologous operations, a CNF naturally takes into consideration both spatial and temporal information to avoid local maxima, which substantially improves the accuracy of motion segmentation. Our motion segmentation system using CNF clearly separates moving objects from their backgrounds. It also effectively eliminates noise to a level achieved when refined post-processing steps are applied to the results of general motion segmentations. It requires less computational power than other random fields and generates automatically normalized outputs without additional post-processes.

Numerous studies have been focusing on the improvement of bag of features (BOF), histogram of oriented gradient (HOG) and scale invariant feature transform (SIFT). However, few works have attempted to learn the connection between them even though the latter two are widely used as local feature descriptor for the former one. Motivated by the resemblance between BOF and HOG/SIFT in the descriptor construction, we improve the performance of HOG/SIFT by a) interpreting HOG/SIFT as a variant of BOF in descriptor construction, and then b) introducing recently proposed approaches of BOF such as locality preservation, data-driven vocabulary, and spatial information preservation into the descriptor construction of HOG/SIFT, which yields the BOF-driven HOG/SIFT. Experimental results show that the BOF-driven HOG/SIFT outperform the original ones in pedestrian detection (for HOG), scene matching and image classification (for SIFT). Our proposed BOF-driven HOG/SIFT can be easily applied as replacements of the original HOG/SIFT in current systems since they are generalized versions of the original ones.

We consider the problem of learning a classifier using only positive and unlabeled samples. In this setting, it is known that a classifier can be successfully learned if the class prior is available. However, in practice, the class prior is unknown and thus must be estimated from data. In this paper, we propose a new method to estimate the class prior by partially matching the class-conditional density of the positive class to the input density. By performing this partial matching in terms of the Pearson divergence, which we estimate directly without density estimation via lower-bound maximization, we can obtain an analytical estimator of the class prior. We further show that an existing class prior estimation method can also be interpreted as performing partial matching under the Pearson divergence, but in an indirect manner. The superiority of our direct class prior estimation method is illustrated on several benchmark datasets.

Most of the recent classification methods require tuning of the hyper-parameters, such as the kernel function parameter and the regularization parameter. Cross-validation or the leave-one-out method is often used for the tuning, however their computational costs are much higher than that of obtaining a classifier. Quadratically constrained maximum a posteriori (QCMAP) classifiers, which are based on the Bayes classification rule, do not have the regularization parameter, and exhibit higher classification accuracy than support vector machine (SVM). In this paper, we propose a multiple kernel learning (MKL) for QCMAP to tune the kernel parameter automatically and improve the classification performance. By introducing MKL, QCMAP has no parameter to be tuned. Experiments show that the proposed classifier has comparable or higher classification performance than conventional MKL classifiers.