An optical flip-flop circuit with a single semiconductor optical amplifier (SOA) using two orthogonal polarization states is proposed. The optical set / reset input and output signals are at a single wavelength. The flip-flop circuit consists of an SOA, a polarization combiner, a polarization splitter, two directional couplers, and two phase shifters. No continuous light source is required to operate the circuit. In this paper, we theoretically analyze the operation performance. Polarization dependence in SOA is considered in the analysis at a single wavelength operation, and numerically simulated results are presented. We confirm that the flip-flop circuit with a feedback-loop length of 15~mm can be operated at switching time of around 3~ns by 1~ns set / reset pulses. The flip-flop performance is discussed from viewpoints of transient overshoot and contrast at the steady on-off states.

Laser diode capable of high speed direct modulation is one of the key solution for short distance applications due to their low power consumption, low cost and small size features. Realization of high modulation bandwidth for direct modulated laser maintaining the above mentioned feature is needed to enhance the short distance, low cost data transmission. One promising approach to enhance the modulation speed is to increase the photon density to achieve high modulation bandwidth. So to achieve this target, 1.55 $mu$m InGaAsP/InGaAsP multiple quantum well (MQW) asymmetric active multimode interferometer laser diode (active MMI-LD) has been demonstrated [1]. The split pumping concept has been applied for the active MMI-LD and significant enhancement of electrical to optical 3 dB down frequency bandwidth (f$_{mathrm{3dB}})$ up to 8 GHz has been successfully confirmed. The reported high bandwidth for split pump active MMI-LD is around 3.5 times higher than the previously reported maximum 3 dB bandwidth (2.3 GHz) of active MMI-LD without split pumping section. That shows, the splitted multimode pumping section behind the electrically isolated modulation section can potentially improve the modulation bandwidth of active MMI-LD. Clear and open eye diagram had also been confirmed for 2.5 Gbps, (2$^{mathrm{7}}$-1) pseudo random bit sequence (PRBS) modulation.

Bit error rate characteristic of negative feedback optical amplifier was investigated by manipulating the negative feedback signal intensity fed into the semiconductor optical amplifier together with the input signal. Consequently, bit error rate was reduced as negative feedback signal intensity increases. Suppression towards the unevenness at the power level `1' and overshoot during rising phase on the output signal eye-diagram was recorded. With negative feedback, through gain decrease of 2.4 dB, power penalty improved remarkably by 15 dB.

Multi-task learning is an important area of machine learning that tries to learn multiple tasks simultaneously to improve the accuracy of each individual task. We propose a new tree-based ensemble multi-task learning method for classification and regression (MT-ExtraTrees), based on Extremely Randomized Trees. MT-ExtraTrees is able to share data between tasks minimizing negative transfer while keeping the ability to learn non-linear solutions and to scale well to large datasets.

The label estimation technique provides a new way to design semi-supervised learning algorithms. If the labels of the unlabeled data can be estimated correctly, the semi-supervised methods can be replaced by the corresponding supervised versions. In this paper, we propose a novel semi-supervised learning algorithm, called Geodesic Weighted Sparse Representation (GWSR), to estimate the labels of the unlabeled data. First, the geodesic distance and geodesic weight are calculated. The geodesic weight is utilized to reconstruct the labeled samples. The Euclidean distance between the reconstructed labeled sample and the unlabeled sample equals the geodesic distance between the original labeled sample and the unlabeled sample. Then, the unlabeled samples are sparsely reconstructed and the sparse reconstruction weight is obtained by minimizing the L1-norm. Finally, the sparse reconstruction weight is utilized to estimate the labels of the unlabeled samples. Experiments on synthetic data and USPS hand-written digit database demonstrate the effectiveness of our method.

This paper describes the design and evaluation of a method for developing a chat-oriented dialog system by utilizing real human-to-human conversation examples from movie scripts and Twitter conversations. The aim of the proposed method is to build a conversational agent that can interact with users in as natural a fashion as possible, while reducing the time requirement for database design and collection. A number of the challenging design issues we faced are described, including (1) constructing an appropriate dialog corpora from raw movie scripts and Twitter data, and (2) developing an multi domain chat-oriented dialog management system which can retrieve a proper system response based on the current user query. To build a dialog corpus, we propose a unit of conversation called a tri-turn (a trigram conversation turn), as well as extraction and semantic similarity analysis techniques to help ensure that the content extracted from raw movie/drama script files forms appropriate dialog-pair (query-response) examples. The constructed dialog corpora are then utilized in a data-driven dialog management system. Here, various approaches are investigated including example-based (EBDM) and response generation using phrase-based statistical machine translation (SMT). In particular, we use two EBDM: syntactic-semantic similarity retrieval and TF-IDF based cosine similarity retrieval. Experiments are conducted to compare and contrast EBDM and SMT approaches in building a chat-oriented dialog system, and we investigate a combined method that addresses the advantages and disadvantages of both approaches. System performance was evaluated based on objective metrics (semantic similarity and cosine similarity) and human subjective evaluation from a small user study. Experimental results show that the proposed filtering approach effectively improve the performance. Furthermore, the results also show that by combing both EBDM and SMT approaches, we could overcome the shortcomings of each.

A twin dominating set of a digraph D is a subset S of vertices if, for every vertex u ∉ S, there are vertices x,y ∈ S such that ux and yu are arcs of D. A digraph D is round if the vertices can be labeled as v0,v1,...,vn-1 so that, for each vertex vi, the out-neighbors of vi appear consecutively following vi and the in-neighbors of vi appear consecutively preceding vi. In this paper, we give polynomial time algorithms for finding a minimum weight twin dominating set and a minimum weight total twin dominating set for a weighted round digraph. Then we show that there is a polynomial time algorithm for deciding whether a locally semicomplete digraph has an independent twin dominating set. The class of locally semicomplete digraphs contains round digraphs as a special case.

In compressed sensing, the design of the measurement matrix is a key work. In order to achieve a more precise reconstruction result, the columns of the measurement matrix should have better orthogonality or linear incoherence. A random matrix, like a Gaussian random matrix (GRM), is commonly adopted as the measurement matrix currently. However, the columns of the random matrix are only statistically-orthogonal. By substituting an orthogonal basis into the random matrix to construct a semi-random measurement matrix and by optimizing the mutual coherence between dictionary columns to approach a theoretical lower bound, the linear incoherence of the measurement matrix can be greatly improved. With this optimization measurement matrix, the signal can be reconstructed from its measures more precisely.

Recognition errors caused by out-of-vocabulary (OOV) words lead critical problems when developing spoken language understanding systems based on automatic speech recognition technology. And automatic vocabulary adaptation is an essential technique to solve these problems. In this paper, we propose a novel and effective automatic vocabulary adaptation method. Our method selects OOV words from relevant documents using combined scores of semantic and acoustic similarities. Using this combined score that reflects both semantic and acoustic aspects, only necessary OOV words can be selected without registering redundant words. In addition, our method estimates probabilities of OOV words using semantic similarity and a class-based N-gram language model. These probabilities will be appropriate since they are estimated by considering both frequencies of OOV words in target speech data and the stable class N-gram probabilities. Experimental results show that our method improves OOV selection accuracy and recognition accuracy of newly registered words in comparison with conventional methods.

This paper studies the design of cooperative beamforming (CB) and cooperative jamming (CJ) for the physical layer security of an amplify-and-forward (AF) relay network in the presence of multiple multi-antenna eavesdroppers. The secrecy rate maximization (SRM) problem of such a network is to maximize the difference of two concave functions, a problem which is non-convex and has no efficient solution. Based on the inner convex approximation (ICA) and semidefinite relaxation (SDR) techniques, we propose two novel low-complexity schemes to design CB and CJ for SRM in the AF network. In the first strategy, relay nodes adopt the CB only to secure transmission. Based on ICA, this design guarantees convergence to a Karush-Kuhn-Tucker (KKT) solution of the SDR of the original problem. In the second strategy, the optimal joint CB and CJ design is studied and the proposed joint design can guarantee convergence to a KKT solution of the original problem. Moreover, in the second strategy, we prove that SDR always has a rank-1 solution for the SRM problem. Simulation results show the superiority of the proposed schemes.

In the process of production design, engineers usually find it is difficult to seek and reuse others' empirical knowledge which is in the forms of lesson-learned documents. This study proposed a novel approach, which uses a semantic-based topic knowledge map system (STKMS) to support timely and precisely lesson-learned documents finding and reusing. The architecture of STKMS is designed, which has five major functional modules: lesson-learned documents pre-processing, topic extraction, topic relation computation, topic weights computation, and topic knowledge map generation modules. Then STKMS implementation is briefly introduced. We have conducted two sets of experiments to evaluate quality of knowledge map and the performance of utilizing STKMS in outfitting design of a ship-building company. The first experiment shows that knowledge maps generated by STKMS are accepted by domain experts from the evaluation since precision and recall are high. The second experiment shows that STKMS-based group outperforms browse-based group in both learning score and satisfaction level, which are two measurements of performance of utilizing STKMS. The promising results confirm the feasibility of STKMS in helping engineers to find needed lesson-learned documents and reuse related knowledge easily and precisely.

We have succeeded in highly selective growth and positioning of Si- and SiGe-quantum-dots (QDs) on SiO2 patterns by controlling the reactive area, whose surface is terminated with OH bonds for Si nucleation in low-pressure chemical vapor deposition (LPCVD). The selective growth of QDs on thermally grown SiO2 line-patterns was demonstrated in LPCVD of SiH4 and GeH4 just after Si nucleation by controlling the early stages of Si2H6-LPCVD, which indicates effectively enhanced initial nucleation on OH-terminated SiO2 surface and suppression of the nucleation and growth of dots on as-grown SiO2 surface during Si2H6-LPCVD prior to SiH4-LPCVD.

In sentiment classification, conventional supervised approaches heavily rely on a large amount of linguistic resources, which are costly to obtain for under-resourced languages. To overcome this scarce resource problem, there exist several methods that exploit graph-based semi-supervised learning (SSL). However, fundamental issues such as controlling label propagation, choosing the initial seeds, selecting edges have barely been studied. Our evaluation on three real datasets demonstrates that manipulating the label propagating behavior and choosing labeled seeds appropriately play a critical role in adopting graph-based SSL approaches for this task.

In this paper, we propose a novel confidence measure to improve the performance of spoken term detection (STD). The proposed confidence measure is based on the context consistency between a hypothesized word and its context in a word lattice. The main contribution of this paper is to compute the context consistency by considering the uncertainty in the results of speech recognition and the effect of topic. To measure the uncertainty of the context, we employ the word occurrence probability, which is obtained through combining the overlapping hypotheses in a word posterior lattice. To handle the effect of topic, we propose a method of topic adaptation. The adaptation method firstly classifies the spoken document according to the topics and then computes the context consistency of the hypothesized word with the topic-specific measure of semantic similarity. Additionally, we apply the topic-specific measure of semantic similarity by two means, and they are performed respectively with the information of the top-1 topic and the mixture of all topics according to topic classification. The experiments conducted on the Hub-4NE Mandarin database show that both the occurrence probability of context word and the topic adaptation are effective for the confidence measure of STD. The proposed confidence measure performs better compared with the one ignoring the uncertainty of the context or the one using a non-topic method.

A reduction in the intensity deviation of a nine-channel optical frequency comb block (OFCB) is demonstrated, by adopting an asymmetric differential drive method for an InP-based dual drive Mach-Zehnder modulator. The generation of a tailored OFCB with an intensity deviation of less than 0.8dB is confirmed by using the modulator.

Web page segmentation has a variety of benefits and potential web applications. Early techniques of web page segmentation are mainly based on machine learning algorithms and rule-based heuristics, which cannot be used for large-scale page segmentation. In this paper, we propose a formulated page segmentation method using visual semantics. Instead of analyzing the visual cues of web pages, this method utilizes three measures to formulate the visual semantics: layout tree is used to recognize the visual similar blocks; seam degree is used to describe how neatly the blocks are arranged; content similarity is used to describe the content coherent degree between blocks. A comparison experiment was done using the VIPS algorithm as a baseline. Experiment results show that the proposed method can divide a Web page into appropriate semantic segments.

“How to get the original ideas” is the fundamental and critical issue for the researchers in science and technology. In this paper, the author writes his experiences concerning how he could encounter the interesting and original ideas of three research subjects, i.e., the accelerating medium effect, the guided-mode extracted integral equation and the surface plasmon gap waveguide.

We propose a novel approach to the target localization problem using Doppler frequency shift measurements. We first reformulate the maximum likelihood estimation (MLE) as a constrained weighted least squares (CWLS) estimation, and then perform the semidefinite relaxation to relax the CWLS problem as a convex semidefinite programming (SDP) problem, which can be efficiently solved using modern convex optimization methods. Finally, the SDP solution can be used to initialize the original MLE which can provide estimates achieve the Cramer-Rao lower bound accuracy. Simulations corroborate the good performance of the proposed method.

Text categorization, especially short text categorization, is a difficult and challenging task since the text data is sparse and multidimensional. In traditional text classification methods, document texts are represented with “Bag of Words (BOW)” text representation schema, which is based on word co-occurrence and has many limitations. In this paper, we mapped document texts to Wikipedia concepts and used the Wikipedia-concept-based document representation method to take the place of traditional BOW model for text classification. In order to overcome the weakness of ignoring the semantic relationships among terms in document representation model and utilize rich semantic knowledge in Wikipedia, we constructed a semantic matrix to enrich Wikipedia-concept-based document representation. Experimental evaluation on five real datasets of long and short text shows that our approach outperforms the traditional BOW method.

This paper describes a robust three-dimensional (3D) surface reconstruction method that can automatically eliminate shadowing errors. For modeling shadowing effect, a new shadowing compensation model based on the angle distribution of backscattered electrons is introduced. Further, it is modified with respect to some practical factors. Moreover, the proposed iterative shadowing compensation method, which performs commutatively between the compensation of image intensities and the modification of the corresponding 3D surface, can effectively provide both an accurate 3D surface and compensated shadowless images after convergence.