Search

This paper proposes a novel approach of weakly supervised video object segmentation, which only needs one pixel to guide the segmentation. We use two deep neural networks to get the instance-level semantic segmentation masks and optical flow maps of each frame. An object probability map to the first frame in video is generated by combining the semantic masks, the optical flow maps and the guiding pixel. The object probability map propagates forward and backward and becomes more accurate to each frame. Finally, an energy minimization problem on a function that consists of unary term of object probability and pairwise terms of label smoothness potentials is solved to get the pixel-wise object segmentation mask of each frame. We evaluate our method on a benchmark dataset, and the experimental results show that the proposed approach achieves impressive performance in comparison with state-of-the-art methods.

A normal human brain holds a high level of bilateral reflection symmetry. On the sagittal view, the brain can be separated into the left and the right hemispheres with approximately identical anatomical properties, so that symmetric mirror pixels have similar properties. As a result, the symmetry information can be used to enhance results of brain segmentation methods. In this paper, we introduced a new version of the Fuzzy C-Mean segmentation method, Symmetry Spatial Possibilistic FCM. SymSPFCM integrates symmetry information with SPFCM which is an extension of PFCM on 3D MR images, and uses the spatial information to calculate possibilistic values and fuzzy membership values. We added spatial and possibilistic information in order to solve the noise sensibility defect of FCM. To integrate the symmetry information, we first extracted the mid-sagittal surface using our proposed method. According to this method, inside each axial slice, a thin-plate spline surface was constructed and local optima were applied to fit this TPS surface to the brain data. Then, the symmetry degree of each symmetry pair was calculated. Finally, the membership values in SPFCM were updated based on the corresponding symmetric values. The efficiency of the proposed method, SymSPFCM, was evaluated using both simulated and real magnetic resonance images and was compared to the state-of-the-art methods. Our results showed images with different degrees of intensity non-uniformity and different levels of noise were segmented efficiently by the SymSPFCM.

Death, injury and disability resulting from road traffic crashes continue to be a major global public health problem. Recent data suggest that the number of fatalities from traffic crashes is in excess of 1.25 million people each year with non-fatal injuries affecting a further 20-50 million people. It is predicted that by 2030 road traffic accidents will have progressed to be the 5th leading cause of death and that the number of people who will die annually from traffic accidents will have doubled from current levels. Both developed and developing countries suffer from the consequences of increase in human population, and therefore, vehicle population. Therefore, methods to reduce accident severity are of great interest to traffic agencies and the public at large. To analyse traffic accident factors effectively we need a complete traffic accident historical database. Any missing data in the database could prevent the discovery of important environmental and road accident factors and lead to invalid conclusions. In this paper, we present a novel imputation method that exploits the within-record and between-record correlations to impute missing data of numerical or categorical values. In addition, our algorithm accounts for uncertainty in real world data by sampling from a list of potential imputed values according to their affinity degree. We evaluated our algorithm using four publicly available traffic accident databases from the United States, the first of which is the largest open federal database (explore.data.gov) in the United States, and the second is based on the National Incident Based Reporting System (NIBRS) of the city and county of Denver (data.opencolorado.org). The other two are from New York's open data portal (Motor Vehicle Crashes-case information: 2011 and Motor Vehicle Crashes-individual information: 2011, data.ny.gov). We compare our algorithm with four state-of-the-art imputation methods using missing value imputation accuracy and RMSE. Our results indicate that the proposed method performs significantly better than the existing algorithms we compared.

Bicluster analysis has emerged as a powerful tool for unsupervised pattern discovery, especially for the analysis of gene expression data. This chapter discusses the biclustering problem, the different bicluster patterns, existing biclustering techniques, and how evolutionary algorithms (EAs) have been applied to solve the biclustering problem. Gene ontology (GO), metabolic pathway maps (MPMs), and protein-protein interaction (PPI) networks can be used to determine the biological functional relevance of genes and conditions in a bicluster. Many biclustering algorithms have been proposed recently and they can be grouped into several categories depending on the bicluster model, the search strategy, and the algorithmic framework used. Factorization-based biclustering algorithm uses spectral decomposition technique to uncover natural substructures that are related to the main patterns of the data matrix. There are two main EA frameworks. The first is based on the genetic algorithm (GA), and the second is based on the artificial immune system (AIS).

Photograph becomes more and more convenient with the rapid growing of smart phones, which makes photo become one of the most widely used social media. In this paper, we have proposed a segmentation and recognition method for multi-model photo event to help users organize and manage their increasing photo collections. Generative model of photo event is built by analyzing time, location, camera parameters and visual content of photos. Expectation-Maximization (EM) learning algorithm is applied to discover the best parameters of the proposed generative model. With the defined model, each photo is categorized into the corresponding event by calculating the maximum posteriori probability. The representativeness of an event is a photo collage constructed by selecting a set of representative photos from the corresponding event. The proposed method is characterized by the following properties: (1) unlike most of the existing photo event segmentation methods, the location of photos is treated as a key feature, (2) the representativeness of an event is a picture collage instead of a single photo, which is not only informative but also very appealing. The experimental results show that the proposed method is effective and efficient on the photo collections from three experienced smart phone users.

In this paper, we propose a combining classifier method based on the Bayesian inference framework. Specifically, the outputs of base classifiers (called Level1 data or meta-data) are utilized in a combiner to produce the final classification. In our ensemble system, each class in the training set induces a distribution on the Level1 data, which is modeled by a multivariate Gaussian distribution. Traditionally, the parameters of the Gaussian are estimated using a maximum likelihood approach. However, maximum likelihood estimation cannot be applied since the covariance matrix of Level1 data of each class is not full rank. Instead, we propose to estimate the multivariate Gaussian distribution of Level1 data of each class by using the Variational Inference method. Experiments conducted on eighteen UCI Machine Learning Repository datasets and a selected 10-class CLEF2009 medical imaging database demonstrated the advantage of our method compared with several well-known ensemble methods. We proposed a Variational Inference method for multivariate Gaussian distribution estimation.We proposed a novel combining classifier method by employing Variational Inference on base classifiers' outputs.The proposed method is highly competitive to many state-of-the-arts ensemble classifier methods.

Designing an ensemble of classifiers is one of the popular research topics in machine learning since it can give better results than using each constituent member. Furthermore, the performance of ensemble can be improved using selection or adaptation. In the former, the optimal set of base classifiers, meta-classifier, original features, or meta-data is selected to obtain a better ensemble than using the entire classifiers and features. In the latter, the base classifiers or combining algorithms working on the outputs of the base classifiers are made to adapt to a particular problem. The adaptation here means that the parameters of these algorithms are trained to be optimal for each problem. In this study, we propose a novel evolving combining algorithm using the adaptation approach for the ensemble systems. Instead of using numerical value when computing the representation for each class, we propose to use the interval-based representation for the class. The optimal value of the representation is found through Particle Swarm Optimization. During classification, a test instance is assigned to the class with the interval-based representation that is closest to the base classifiers’ prediction. Experiments conducted on a number of popular dataset confirmed that the proposed method is better than the well-known ensemble systems using Decision Template and Sum Rule as combiner, L2-loss Linear Support Vector Machine, Multiple Layer Neural Network, and the ensemble selection methods based on GA-Meta-data, META-DES, and ACO.

Ensemble selection is one of the most studied topics in ensemble learning because a selected subset of base classifiers may perform better than the whole ensemble system. In recent years, a great many ensemble selection methods have been introduced. However, many of these lack flexibility: either a fixed subset of classifiers is pre-selected for all test samples (static approach), or the selection of classifiers depends upon the performance of techniques that define the region of competence (dynamic approach). In this paper, we propose an ensemble selection method that takes into account each base classifier's confidence during classification and the overall credibility of the base classifier in the ensemble. In other words, a base classifier is selected to predict for a test sample if the confidence in its prediction is higher than its credibility threshold. The credibility thresholds of the base classifiers are found by minimizing the empirical 0–1 loss on the entire training observations. In this way, our approach integrates both the static and dynamic aspects of ensemble selection. Experiments on 62 datasets demonstrate that the proposed method achieves much better performance in comparison to some ensemble methods.

Many image processing and pattern recognition problems can be formulated as binary quadratic programming (BQP) problems. However, solving a large BQP problem with a good quality solution and low computational time is still a challenging unsolved problem. Current methodologies either adopt an independent random search in a semi-definite space or perform search in a relaxed biconvex space. However, the independent search has great computation cost as many different trials are needed to get a good solution. The biconvex search only searches the solution in a local convex ball, which can be a local optimal solution. In this paper, we propose a BQP solver that alternatingly applies a deterministic search and a stochastic neighborhood search. The deterministic search iteratively improves the solution quality until it satisfies the KKT optimality conditions. The stochastic search performs bootstrapping sampling to the objective function constructed from the potential solution to find a stochastic neighborhood vector. These two steps are repeated until the obtained solution is better than many of its stochastic neighborhood vectors. We compare the proposed solver with several state-of-the-art methods for a range of image processing and pattern recognition problems. Experimental results showed that the proposed solver not only outperformed them in solution quality but also with the lowest computational complexity.

With the wide availability of the Internet and the proliferation of pornographic images online, adult image detection and filtering has become very important to prevent young people from reaching these harmful contents. However, due to the large diversity in adult images, automatic adult image detection is a difficult task. In this paper, a new deep convolutional neural network (DCNN) based approach is proposed to classify images into three classes, i.e. porn, sexy, and benign. Our approach takes both the entire picture (global context) and the meaningful region (local context) information into consideration. The proposed network is composed of three parts, i.e. the image characteristics subnet to extract discriminative low-level image features, the sensitive body part detection subnet to detect adult-image related regions, and the feature extraction and fusion subnet to generate high-level features for image classification. A multi-task learning scheme is designed to optimize the network with both the global and local information. Experiments are carried out on two datasets with over 160,000 images. From the experiment results, it was observed that the proposed network achieved high classification accuracies (96.6% in the AIC dataset and 92.7% in the NPDI dataset) and outperformed the other approaches investigated.

Malonylation is a recently discovered post-translational modification (PTM) in which a malonyl group attaches to a lysine (K) amino acid residue of a protein. In this work, a novel machine learning model, SPRINT-Mal, is developed to predict malonylation sites by employing sequence and predicted structural features. Evolutionary information and physicochemical properties are found to be the two most discriminative features whereas a structural feature called half-sphere exposure provides additional improvement to the prediction performance. SPRINT-Mal trained on mouse data yields robust performance for 10-fold cross validation and independent test set with Area Under the Curve (AUC) values of 0.74 and 0.76 and Matthews' Correlation Coefficient (MCC) of 0.213 and 0.20, respectively. Moreover, SPRINT-Mal achieved comparable performance when testing on H. sapiens proteins without species-specific training but not in bacterium S. erythraea. This suggests similar underlying physicochemical mechanisms between mouse and human but not between mouse and bacterium. SPRINT-Mal is freely available as an online server at: http://sparks-lab.org/server/SPRINT-Mal/. © 2018 Wiley Periodicals, Inc.

In this paper, we propose a novel method for efficient boundary detection in close-range hyperspectral images (HSI). We adopt different spectral similarity measurements to construct a sparse spectral-spatial affinity matrix that characterizes the similarity between the spectral responses of neighboring pixels within a local neighborhood. After that, we adopt a spectral clustering method in which the eigenproblem is solved and the eigenvectors of smallest eigenvalues are calculated. Morphological erosion is then applied on each eigenvector to detect the boundary. We fuse the results of all eigenvectors to obtain the final boundary map. Our method is evaluated on a real-world HSI dataset and compared with three alternative methods. The results exhibit that our method outperforms the alternatives, and can cope with several scenarios that methods based on color images can not handle.

Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved Biclustering is a powerful unsupervised learning technique that has different applications in many fields especially in gene expression analysis. This technique tries to group rows and columns in a dataset simultaneously, which is an NP-hard problem. In this paper, a multi-objective evolutionary algorithm is proposed with a heuristic search to solve the biclustering problem. To do so, rows are projected into the column space. Projection decreases the computational cost of geometric biclustering. The heuristic search is done by sample Pearson correlation coefficient over the rows and columns of a dataset to prune unwanted rows and columns. The experimental results on both synthetic and real datasets show the effectiveness of our proposed method.

In this paper, we address the image region tagging procedure in which each image region is annotated by a suitable concept. Specifically, we first extract the feature vector for each segmented region. Then we propose a Genetic Algorithm (GA)-based simultaneous classifier and feature selection method working with ensemble system to learn the relationship between the low-level features and high-level concepts. The extensive experiments conducted on two public datasets namely MSRC v1 and MSRC v2 demonstrate the better performance of our method than several well-known ensemble methods, supervised machine learning methods, and sparse coding-based methods in the regions-in-image classification task.

In this study, we introduce an ensemble-based approach for online machine learning. Here, instead of working on the original data, several Hoeffding tree classifiers classify and are updated on the lower dimensional projected data generated from originality by random projections. Since random projection is unstable, from one example, many diverse training data can be created to train the set of Hoeffding tree classifiers. The experiments conducted on a number of datasets chosen from different sources demonstrate that the proposed approach performs significantly better than the single Hoeffding tree and some well-known online learning algorithms including additive models and Online Bagging.

Boundary detection in hyperspectral image (HSI) is a challenging task due to high data dimensionality and the that is distributed over the spectral bands. For this reason, there is a dearth of research on boundary detection in HSI. In this paper, we propose a spectral-spatial feature based statistical co-occurrence method for this task. We adopt probability density function (PDF) to estimate the co-occurrence of features at neighboring pixel pairs. Such cooccurrence is rare at the boundary and repeated within a region. To fully explore the material information embedded in HSI, joint spectral-spatial features are extracted at each pixel. The PDF values are then used to construct an affinity matrix for all pixels. After that, a spectral clustering algorithm is applied on the affinity matrix to produce boundaries. Our algorithm is evaluated on a dataset of real-world HSIs and compared with two alternative approaches. The results show that the proposed method is very effective in exploring object boundaries from HSI images.

We present OCSB, a novel online Bayesian framework for imbalance multi-class data streams. To the best of our knowledge, OCSB is the first online method applying both cost-sensitive learning and sampling technique in a single classifier to deal with class imbalance learning. Specifically, an artificial cost matrix is designed and adapted in a sequential manner to not only boost the accuracy of minority classes but also guarantee stable Gmean - geometric mean of the accuracies for all classes. Furthermore, we introduce a new intermediate random sampling strategy with over- sampled minority classes and under-sampled majority classes. This offers twofold benefit: learning the rare classes properly and reducing the cost caused by the redundant data of majority classes. Experimental results show that our OCSB outperforms very recent well-known methods for online imbalanced learning algorithms in the literature.

Digital curation is an emerging topic of interest in museum studies. On-site museum experience enhanced by digital technology can not only promote the social and educational roles of traditional museums in the digital era but also boost the value of cultural heritage. Our objective is to design an on-site personalized museum recommendation system to extend the connection between visitors and collections, focusing on the user experience. The prototype is made up of two different functional components, a mobile user interface (UI) at the front-end for visitors to interact with the collections during their visit, and an object database in the back-end where information is stored and accessed using semantic technology. In this article, we discuss the design of the UI for the collection of Australian paintings in the National Gallery of Australia, and evaluate the framework for visitor behaviour capture and analysis using the interface, thereby proposing a system for establishing a personalized tour trajectory.

View-invariant gait recognition is one of the major challenges in identifying people through their gait. Many researchers have evaluated view angle transformation techniques, discriminant analysis and manifold learning approaches for cross-view recognition, and their proposals are usually based on a common factor, i.e., to establish a cross-view mapping between gallery and probe templates. However, their effectiveness is restricted to small view angle variances. A promising approach to perform view-invariant gait recognition is through multi-view feature learning. In this paper, we propose the view-invariant feature selector (ViFS) and integrate it in a framework for view-invariant gait recognition. ViFS select features from multi-view gait templates and reconstructs gallery templates that accurately match the data for a specific view angle. ViFS is thus able to reconstruct gallery templates from arbitrary view angles, and thus help to transfer the cross-view problem to identical-view gait recognition. We also apply linear subspace learning methods as feature enhancers for ViFS, which substantially reduce the computational cost and improve the recognition speed. We test the proposed framework on the CASIA Dataset B. The average recognition accuracy of the proposed framework for 11 different views exceed 98%.

Hand posture estimation is an important step in hand gesture detection. It refers to the process of modeling hand in computer to accurately represent the actual hand obtained from an acquisition device. In the literature, several objective functions (mostly based on silhouette or point cloud) have been used to formulate and solve the problem of hand posture estimation as a minimisation problem using stochastic or deterministic algorithms. The main challenge is that the objective function is computationally expensive. In the case of using point clouds, decreasing the number of points results in a better computational cost, but it decreases the accuracy of hand posture estimation. We argue in this paper that hand posture estimation is a bi-objective problem with two conflicting objectives: minimising the error versus minimising the number of points in the point cloud. As an early effort, this paper first formulates hand posture estimation as a bi-objective optimisation problem and then approximates its true Pareto optimal front with an improved Multi-Objective Particle Swarm Optimisation (MOPSO) algorithm. The proposed algorithm is used to determine the Pareto optimal front for 16 hand postures and compared with the original MOPSO. The results proved that the objectives are in conflict and the improved MOPSO outperforms the original algorithm when solving this problem.

Background Cerebral structural changes related to obstructive sleep apnoea (OSA) have been reported in adult OSA patients; however, similar data and their associations with neurocognitive dysfunction are scarce in childhood OSA. Objective To compare neurocognitive function, regional grey matter density and cerebral volume in children with and without OSA. Methods Fifty OSA cases and 27 normal controls underwent a panel of neurocognitive tests. High resolution 3-dimensional magnetic resonance images of the brain were obtained from 23 OSA cases and 15 gender and age matched controls. Total cerebral volume and regional grey matter density were analyzed using voxel-based morphometry technique and compared between the two groups. Individuals with an obstructive apnoea hypopnoea index (OAHI) > 5 were defined as having moderate-to-severe OSA. Results Children with OSA showed significantly reduced attention and visual-fine motor coordination scores compared with controls. Grey matter volume deficit was observed in prefrontal and temporal regions of cases with moderate-to-severe OSA only. Significant negative correlations were found between the visual-fine motor coordination score and the ratio of grey matter volume over total brain volume. Conclusion Children with OSA had impaired attention and visual-fine motor coordination. Regional grey matter reduction was evident in children with more severe OSA.

The Wiener index is the most popular topological index used as an invariant molecular descriptor in the study of structures and physiochemical properties of molecules. Here, Wiener index and its modifications, hyper-Wiener and Harary indices, as well as Hosoya and hyper-Hosoya polynomials are studied over C 84 fullerene nanocage which is the third most abundant member of the higher fullerene series with two major structures known as stable C84, D2 (22) and D2d (23). The computation of Wiener indices and polynomials of isomers C84–D2 (22) and C84–D2d (23) is implemented using graph algorithms which employ the isomer’s molecular graphs. Our results show that the isomer C84–D2 (22) has greater Wiener and topological efficiency indices than isomer C84-D2d (23), which indicates that the isomer C84–D2d (23) has greater compactness and stability than isomer C84–D2 (22). These results are consistent with the experimental observations and further confirmed by the lower Harary index of isomer C84–D2 (22). The greater Wiener index of isomer C84–D2 (22) also suggests that its boiling point would be higher than that of isomer C84–D2d (23).

Traditional image aesthetic evaluation method usually involves the extraction of a set of relevant image aesthetic features and classification by a classifier trained on the set of features. The system's performance greatly depends on the effectiveness of the features. However, most of these features are carefully hand-crafted for specific datasets and assumed strong prior knowledge. Therefore, these features would not be optimal for general image aesthetic evaluation. The deep convolution neural network (DCNN) has the ability to automatically learn aesthetic features, and network structure of different complexity can learn aesthetic features at different scales and different point of views. Moreover, traditional image features, such as edge and saliency map, can be used as auxiliary information for the DCNN. Therefore, a Network-Paralleled and Data-Paralleled DCNN (NP-DP-DCNN) structure is proposed. The Network-Paralleled DCNN fuses networks of different complexity and the Data-Paralleled DCNN fuses original image data and derived feature maps to learn the aesthetic features from different scales and different point of views. Experimental results show that the proposed NP-DP-DCNN structure is able to achieve better classification performance than many existing methods.

Compared with deep neural network which is trained using back propagation, the extreme learning machine (ELM) learns thousands of times faster but still produces good generalization performance. To better understand the ELM, this paper studies the effect of noise on the input nodes or hidden neurons. It was found that there is no effect on the performance of ELM when small amount of noise is added to the input or the neurons in the hidden layer. Although the performance of ELM would improve with an increase in the number of neurons in the hidden layer, beyond a certain limit, this could lead to overfitting. In view of this, a parallel ELM (P-ELM) is proposed to improve the system performance. P-ELM has better robustness to noise due to the ensemble nature and is less susceptible to overfitting since each parallel hidden layer has only a moderate number of hidden neurons. Experimental results have indicated that the proposed P-ELM can achieve better classification performance than ELM without large increase in training time.

Data stream mining has been receiving increased attention due to its presence in a wide range of applications, such as sensor networks, banking, and telecommunication. One of the most important challenges during learning from data streams is reacting to concept drift, i.e. unforeseen changes of the stream's underlying data distribution. Traditional methods always used online learning to handle the concept drift problem. However, online learning requires high time cost during online training. To overcome this shortcoming, this paper proposes a Kalman filtering approach, which can provide robust concept drift detection, to track concept drift. Once concept drift happens, the online extreme learning machine is applied to update the tracking model, whereas the offline extreme learning machine is used when no concept drift occurs. Based on this idea, we propose a fusion framework to combine online and offline extreme learning machine to efficiently track the data stream. The experiment results indicate the superior performance of our method.

We present VIGO, a novel online Bayesian classifier for both binary or multiclass problems. In our model, variational inference for multivariate Gaussian distribution technique is exploited to approximate the class conditional probability density functions of data in an online manner. Besides being a conservative learner with a low number of updates compared with many other popular algorithms, VIGO algorithm can be updated in a minibatch of an arbitrary size which makes it robust with noise data. Experiments over a large number of UCI datasets demonstrate the advantage of VIGO with many state-of-the-art methods presented in LIBOL - a prevalent library for online learning algorithms.

Research has shown that the human lip and its movements are a rich source of information related to speech content and speaker's identity. Lip image segmentation, as a fundamental step in many lip-reading and visual speaker authentication systems, is of vital importance. Because of variations in lip color, lighting conditions and especially the complex appearance of an open mouth, accurate lip region segmentation is still a challenging task. To address this problem, this article proposes a new fuzzy deep neural network having an architecture that integrates fuzzy units and traditional convolutional units. The convolutional units are used to extract discriminative features at different scales to provide comprehensive information for pixel-level lip segmentation. The fuzzy logic modules are employed to handle various kinds of uncertainties and to provide a more robust segmentation result. An end-to-end training scheme is then used to learn the optimal parameters for both the fuzzy and the convolutional units. A dataset containing more than 48 000 images of various speakers, under different lighting conditions, was used to evaluate lip segmentation performance. According to the experimental results, the proposed method achieves state-of-the-art performance when compared with other algorithms.

Motivation: Kinesin is a smart motor protein that steps processively forward and backward along microtubules (MTs). The mechanical kinetics of kinesin affecting its stepping behavior is not fully understood. Here, we propose a mathematical model to study the mechanical kinetics of forward and backward stepping of kinesin motor based on the four-state discrete stochastic model of the motor. Results: Results show that the probabilities of forward and backward stepping can be modeled using the mean probabilities of forward and backward kinetic transitions, respectively. We show that the backward stepping of kinesin motor starts when the probability of adenosine diphosphate (ADP) binding to the motor is much higher than that of adenosine triphosphate (ATP) binding. Furthermore, our results indicate that the backward stepping is related to both ATP hydrolysis and synthesis with rate limiting factor being ATP synthesis. Low rate of ATP synthesis under high backward loads above 10 pN is also suggested as a reason for the detachment of kinesin motor from MT in the kinetic state MTċKinesinċADPċPi. Availability and implementation: The code for this work is written in Visual C# and is available by request from the authors. Contact: a.liew@griffith.edu.au

Kinesin and muscle myosin are considered as physical bio-nanoagents able to sense their cells through their sensors, make decision internally, and perform actions through their actuators. This paper has investigated and compared the flexible (reactive, pro-active, and interactive) autonomous behaviors of kinesin and muscle myosin bio-nanorobots. Using an automata algorithm, the agent-based deterministic finite automaton models of the internal decision making processes of the bio-nanorobots (as their reactive and pro-active capabilities) were converted to their respective computational regular languages (as their interactive capabilities). The resulted computational languages could represent the flexible autonomous behaviors of the bio-nanorobots. The proposed regular languages also reflected the degree of the autonomy and intelligence of internal decision-making processes of the bio-nanorobots in response to their environments. The comparison of flexible autonomous behaviors of kinesin and muscle myosin bio-nanorobots indicated that both bio-nanorobots employed regular languages to interact with their environments through two sensors and one actuator. Moreover, the results showed that kinesin bio-nanorobot used a more complex regular language to interact with its environment compared with muscle myosin bio-nanorobot. Therefore, our results have revealed that the flexible autonomous behavior of kinesin bio-nanorobot was more complicated than the flexible autonomous behavior of muscle myosin bio-nanorobot.

Computational nanotechnology provides theoretical understanding of complex nanoparticles properties and may enlighten their potential applications. In this paper, we use graph algorithms based on the dynamic programming to develop a software system which computes a set of optimal physical properties of the C20 nanoparticle. The computational results of our developed software system have not been reported previously. C20 nanoparticle is the smallest and simplest fullerene composed of twenty equivalent carbon atoms arranged in a cage structural form. The twenty carbon atoms of C20 nanoparticle are also arranged in two other major structures: bowl and ring. The study considers all three major structures of C20 nanoparticle and computes their physical indices including Wiener, hyper-Wiener, Harary and reciprocal Wiener. The Hosoya and hyper-Hosoya polynomials of the cage, bowl and ring structures of C20 nanoparticle have been also computed. The accuracy of the computational results was proved with the mathematical equations between the indices and polynomials. Our computational results also showed good agreement with the physical properties of the cage, bowl and ring structures of C20 nanoparticle.

Sensors have been developed and applied in a wide range of fields such as robotics and autonomous vehicle navigation (AVN). Due to the inability of a single sensor to fully sense its surroundings, multiple sensors based on individual specialties are commonly used in order to complement the shortcomings and enrich perception. However, it is challenging to integrate the heterogeneous types of sensory information and produce useful results. This research aims to achieve a high degree of accuracy with a minimum false-positive and false-negative rate for the sake of reliability and safety. This paper introduces a likelihood-based data fusion model, which integrates information from various sensors, maps it into the integrated data space and generates the solution considering all the information from the sensors. Two distinct sensors: an optical camera and a LIght Detection And Range (Lidar) sensor were used for the experiment. The experimental results showed the usefulness of the proposed model in comparison with single sensor outcomes.

This paper proposes a method that uses both spectral and spatial information to segment remote sensing hyperspectral images. After a hyperspectral image is over-segmented into superpixels, a deep Convolutional Neural Network (CNN) is used to perform superpixel-level labelling. To further delineate objects from a hyperspectral scene, this paper attempts to combine the properties of CNN and Conditional Random Field (CRF). A mean-field approximation algorithm for CRF inference is used and formulated with Gaussian pairwise potentials as Recurrent Neural Network. This combined network is then plugged into the CNN which leads to a deep network that has robust characteristics of both CNN and CRF. Preliminary results suggest the usefulness of this framework to a promising extent.

Biclustering has many applications in various fields such as pattern classification, information retrieval, data mining and functional annotation. Biclustering extracts accurate information from gene expression datasets by clustering rows and columns of a dataset simultaneously. In this paper, a new multi-objective evolutionary biclustering framework based on strength Pareto front evolutionary algorithm (SPEA2) is proposed. A heuristic search is added into SPEA2 to delete and add genes and conditions into randomly generated biclusters. In order to select the best bicluster among Pareto front solutions, k-mean algorithm is used. The new population is generated based on mutation and crossover. The performance of the proposed method is evaluated using synthetic and real datasets and compared with several well-known biclustering methods. The experimental results show better performance and significant enrichment of detected biclusters.

Compared with other traditional biometric features such as face, fingerprint, or handwriting, lip biometric features contain both physiological and behavioral information. Physiologically, different people have different lips. On the other hand, people can usually be differentiated by their talking style. Current research on lip biometrics generally does not distinguish between the two kinds of information during feature extraction and classification and the interesting question of whether the physiological or the behavioral lip features are more discriminative has not been comprehensively studied. In this paper, different physiological and behavioral lip features are studied with respect to their discriminative power in speaker identification and verification. Our experimental results have shown that both the static lip texture feature and the dynamic shape deformation feature can achieve high identification accuracy (above 90%) and low verification error rate (below 5%). In addition, the lip rotation and centroid deformations, which are related to the speaker's talking mannerism, are found to be useful for speaker identification and verification. In contrast to previous studies, our results show that behavioral lip features are more discriminative in speaker identification and verification compared to physiological features.

Biclustering analysis is a useful methodology to discover the local coherent patterns hidden in a data matrix. Unlike the traditional clustering procedure, which searches for groups of coherent patterns using the entire feature set, biclustering performs simultaneous pattern classification in both row and column directions in a data matrix. The technique has found useful applications in many fields but notably in bioinformatics. In this paper, we give an overview of the biclustering problem and review some existing biclustering algorithms in terms of their underlying methodology, search strategy, detected bicluster patterns, and validation strategies. Moreover, we show that geometry of biclustering patterns can be used to solve biclustering problems effectively. Well-known methods in signal and image analysis, such as the Hough transform and relaxation labeling, can be employed to detect the geometrical biclustering patterns. We present performance evaluation results for several of the well known biclustering algorithms, on both artificial and real gene expression datasets. Finally, several interesting applications of biclustering are discussed.

Protein-peptide interactions are essential for all cellular processes including DNA repair, replication, gene-expression, and metabolism. As most protein-peptide interactions are uncharacterized, it is cost effective to investigate them computationally as the first step. All existing approaches for predicting protein-peptide binding sites, however, are based on protein structures despite the fact that the structures for most proteins are not yet solved. This article proposes the first machine-learning method called SPRINT to make Sequence-based prediction of Protein-peptide Residue-level Interactions. SPRINT yields a robust and consistent performance for 10-fold cross validations and independent test. The most important feature is evolution-generated sequence profiles. For the test set (1056 binding and non-binding residues), it yields a Matthews' Correlation Coefficient of 0.326 with a sensitivity of 64% and a specificity of 68%. This sequence-based technique shows comparable or more accurate than structure-based methods for peptide-binding site prediction. SPRINT is available as an online server at: http://sparks-lab.org/. © 2016 Wiley Periodicals, Inc.

Fluorescence microscopy image segmentation is a central task in high-throughput applications such as protein expression quantification and cell function investigation. In this paper, a multiple kernel local level set segmentation algorithm is introduced as a framework for fluorescence microscopy cell image segmentation. In this framework, a new local region-based active contour model in a variational level set formulation based on the piecewise constant model and multiple kernels mapping is proposed where a linear combination of multiple kernels is utilized to implicitly map the original local image data into data of a higher dimension. We evaluate the performance of the proposed method using a large number of fluorescence microscopy images. A quantitative comparison is also performed with some state-of-the-art segmentation approaches.

With microarray technology, gene expression profiles are produced at a rapid rate. It remains a challenge for biologists to robustly identify periodic gene expression profiles when the time series have short data length and contain a high level of noise. An effective method is proposed in this paper to analyze the periodicity of gene expression time series us- ing singular value decomposition (SVD), singular spectrum analysis (SSA) and autoregressive (AR) model-based spec- tral estimation. Using these procedures, noise can be filtered out and over 85% of periodic gene expression can be identi- fied in the mouse segmentation clock data set. AMS 2000 subject classifications: Primary 60K35, 60K35; secondary 60K35. Keywords and phrases: Singular value decomposition (SVD), Singular spectrum analysis (SSA), Segmentation clock, Periodicity analysis, Microarray time series analysis. Transcription profiling of mouse presomitic mesoderm with 17 samples at different time points is carried out to iden- tify periodic genes of the segmentation clock (1). Based on this dataset, Dequeant (3) carried out a research study com- paring the pattern detection performance of several math- ematical approaches, which included the Lomb-Scargle (L) periodogram, Phase consistency (P), Address reduction (A), Cyclohedron test (C), and Stable persistence (S). The top three hundred ranked probe sets from these five methods were found and the results show that the Stable persistence (S) method performs best by identifying most of the bench- mark probe sets within the top 300 probe sets. However, the data contain a high level of noise, which will degrade the performance of most data analysis algorithms. There- fore, we need to develop an effective method to process the noisy time series data. In this paper, an effective method is developed to iden- tify the periodicity of microarray time series data by com- bining singular value decomposition (SVD), singular spec- trum analysis (SSA) and autoregressive (AR) model-based spectral analysis. By considering the singular values of time series data, the noise can be reduced (4). By using AR mod- eling, more accurate spectral estimation results are obtained (5). In our work, about 85% of gene expression profiles in the mouse segmentation clock dataset are found to be periodic. 2. METHODS

Many cellular processes exhibit periodic behaviors. Hence, one of the important tasks in gene expression data analysis is to detect subset of genes that exhibit cyclicity or periodicity in their gene expression time series profiles. Unfortunately, gene expression time series profiles are usually of very short length, with very few periods, irregularly sampled and are highly contaminated with noise. This makes the detection of periodic profiles a very challenging problem. Recently, a hypothesis testing method based on the Fisher g-statistic with correction for multiple testing has been proposed to detect periodic gene expression profiles. However, it was observed that the test is not reliable if the signal length is too short. In this paper, we performed extensive simulation study to investigate the statistical power of the test as a function of noise distribution, signal length, SNR, and the false discovery rate (FDR). We have found that the number of periodic profiles can be severely underestimated for short length signal. The findings indicate that caution needs to be exercised when interpreting the test result for very short length signals.

It is well known that visual cues of lip movement contain important speech relevant information. This paper presents an automatic lipreading system for small vocabulary speech recognition tasks. Using the lip segmentation and modeling techniques we developed earlier, we obtain a visual feature vector composed of outer and inner mouth features from the lip image sequence for recognition. A spline representation is employed to transform the discrete-time sampled features from the video frames into the continuous domain. The spline coefficients in the same word class are constrained to have similar expression and are estimated from the training data by the EM algorithm. For the multiple-speaker/speaker-independent recognition task, an adaptive multimodel approach is proposed to handle the variations caused by various talking styles. After building the appropriate word models from the spline coefficients, a maximum likelihood classification approach is taken for the recognition. Lip image sequences of English digits from 0 to 9 have been collected for the recognition test. Two widely used classification methods, HMM and RDA, have been adopted for comparison and the results demonstrate that the proposed algorithm deliver the best performance among these methods.

Robust and accurate lip region segmentation is of vital importance for lip image analysis. However, most of the current techniques break down in the presence of mustaches and beards. With mustaches and beards, the background region becomes complex and inhomogeneous. We propose in this paper a novel multi-class, shape-guided FCM (MS-FCM) clustering algorithm to solve this problem. For this new approach, one cluster is set for the object, i.e. the lip region, and a combination of multiple clusters for the background which generally includes the skin region, lip shadow or beards. The proper number of background clusters is derived automatically which maximizes a cluster validity index. A spatial penalty term considering the spatial location information is introduced and incorporated into the objective function such that pixels having similar color but located in different regions can be differentiated. This facilitates the separation of lip and background pixels that otherwise are inseparable due to the similarity in color. Experimental results show that the proposed algorithm provides accurate lip-background partition even for the images with complex background features like mustaches and beards.

Microarray data clustering has drawn great attention in recent years. However, a major problem in data clustering is convergence to a local optimal solution. In this paper, we introduce a regularized version of the l 2m-FCM algorithm to resolve this problem. The strategy is to constrain the descent direction in the optimization procedure. For this we employ a novel method, calculus of variations, to correct the direction. Experimental results show that the proposed method has a better performance than seven other clustering algorithms for three synthetic and six real world data sets. Also, the proposed method produces reliable results for synthetic data sets with a large number of groups, which is a challenging problem for many clustering algorithms. Our method has been applied to microarray data classification with good results.

Hybrid broadcast and broadband (HBB) television is a new platform that opens many possibilities for new services. Recommendation system offers a personalized service that suggests items of interest according to user preference. Nowadays, the number of available programs is so large that one cannot realistically have a real time overview. Recommendation engines were developed to solve the problem of information overload, and save time and effort when looking for appealing content. In this paper, we present model design and implementation of a recommendation system for HBB TV. To explore user preferences and make predictions, an enhanced Naive Bayes model for rating prediction is designed. The model uses a set of features to predict user rating based on past observation. The recommendation system presented in this paper is flexible and robust enough to handle a sparse data set with very few records of feature description. Experiments performed on a Yahoo movie data set indicated the promising performance of our approach.

The amount of available content of different types of services is so large nowadays that one cannot realistically have a real time overview of the content. Recommendation engines were developed to solve the problem of information overload, and save time and effort when looking for appealing content. In this paper, we present an enhanced Naive Bayes model for rating prediction of a program based on content description information. As our prediction model has to deal with categorical data, a probabilistic Bayesian network is used. The model uses a set of features to predict user rating based on past observation. We also simulated recommendation from a program offer. The recommendation system presented in this paper is flexible and robust enough to handle a sparse data set with very few records of feature description. Experiments were performed on a Yahoo movie data set and they indicated the promising performance of our approach over an existing technique.

Most art paintings especially abstract art are created to evoke an emotional feeling in the viewer. Colour as a dominant low-level feature plays a significant role in visual perception. Colours can bring out different emotional feelings and the findings from several psychology studies have shown bright colours generally relate to positive emotions, whereas dark colours relate to negative emotions. This paper focuses on the study of relationships between colours and emotionson abstract art images. In this work, the colour featurevectors were extracted from images using the fuzzy colour histogram method, whereas emotion typevectors were obtained from a user rating study done on an abstract art imagedataset. Ourstudy shows that some emotion types have highly significant association with certain colours. For example, the fear emotion has a positive correlation with the colour black with a significant value of p < 0.01.

Clustering is an unsupervised learning technique that groups data into clusters using the entire conditions. However, sometimes, data is similar only under a subsetof conditions. Biclustering allows clustering of rows and columns of a dataset simultaneously. It extracts more accurate information from sparse datasets. In recent years, biclustering has found many useful applications in different fields and many biclustering algorithms have been proposed. Using both row and column information of data, biclustering requires the optimization of two conflicting objectives. In this study, a new multi-objective evolutionary biclustering framework using SPEA2 is proposed. A heuristic local search based on the gene and condition deletion and addition is added into SPEA2 and the best bicluster is selected usinga new quantitative measure that considers both its coherence and size. The performance of our algorithm is evaluatedusing simulated and gene expression data and compared with several well-known biclustering methods. The experimental results demonstrate better performance with respect to the size and MSR of detected biclusters and significant enrichment of detected genes.