Search

Citation recommendation can help researchers quickly find supplementary or alternative references in massive academic resources. Current research on citation recommendation mainly focuses on the citing papers, resulting in the enormous cited papers are ignored, including the relations among cited papers and their citation context cited in citing papers. Moreover, cited paper’s content is often denoted with its original title and abstract, which is hard to acquire and rarely considers different citation motivations. Furthermore, the most appropriate method for semantic representation of cited papers’ relations and content is uncertain. Therefore, this paper studies citation recommendation from the perspective of semantic representation of cited papers’ relations and content. Firstly, four forms of citation context are designed and extracted as cited papers’ content considering citation motivations, as well as co-citation relationships are extracted as cited papers’ relations. Secondly, 132 methods are designed for generating semantic vector of cited paper, including four network embedding methods, 16 methods by combining four text representation algorithms with four forms of citation content, and 112 fusion methods. Finally, similarity among cited papers is calculated for citation recommendation and a quantitative evaluation method based on link prediction is designed, to find the most appropriate form of citation content and the optimal method. The result shows that doc2vecC (Document to Vector through Corruption) with the form of CS&SS (Current Sentences and Surrounding Sentences) performs best, in which the AUC (Area Under Curve) and MAP (Macro Average Precision) reach 0.877 and 0.889 and have increased by 0.462 and 0.370 compared with the worst-performing method. This performance is slightly improved by parameters adjustment, and a case study is performed whose results have further proved the effectiveness of this method. In addition, among four forms of cited papers’ content, CS&SS performs best in almost all methods. Furthermore, the fusion methods not always perform better than the single methods, where doc2vecC (CS&SS) performs better than the best fusion method GCN (Graph Convolutional Network). These results not only prove the effectiveness of citation recommendation from the perspective of cited paper, but also provide helpful and useful suggestions for method selection and citation content selection. The data and conclusions can be extended to other text mining-related tasks. Simultaneously, it is a preliminary research which needs to be further studied in other domains using emerging semantic representation methods.

Literature reviews are essential for any researcher trying to keep up to date with the burgeoning software engineering literature. FAST$^2$ is a novel tool for reducing the effort required for conducting literature reviews by assisting the researchers to find the next promising paper to read (among a set of unread papers). This paper describes FAST$^2$ and tests it on four large software engineering literature reviews conducted by Wahono (2015), Hall (2012), Radjenovi\'c (2013) and Kitchenham (2017). We find that FAST$^2$ is a faster and robust tool to assist researcher finding relevant SE papers which can compensate for the errors made by humans during the review process. The effectiveness of FAST$^2$ can be attributed to three key innovations: (1) a novel way of applying external domain knowledge (a simple two or three keyword search) to guide the initial selection of papers---which helps to find relevant research papers faster with less variances; (2) an estimator of the number of remaining relevant papers yet to be found---which in practical settings can be used to decide if the reviewing process needs to be terminated; (3) a novel self-correcting classification algorithm---automatically corrects itself, in cases where the researcher wrongly classifies a paper., Comment: 20+3 pages, 6 figures, 5 tables, and 4 algorithms. Accepted by Journal of Expert Systems with Applications

Paper is an example of a valuable material that can be recycled and recovered. In this study, a two-stage stochastic revenue-maximization model is presented to determine a long-term strategy under uncertainty for a large-scale real-world paper recycling company. This network-design problem includes optimal recycling center locations and optimal flow amounts between the nodes in the multi-facility environment. The proposed model is formulated with two-stage stochastic mixed-integer and robust programming approaches. The models are solved by commercial software GAMS 21.6/CPLEX 9.0 and the results are compared. The study is followed by the analyses of the results.

This article explores the application of image processing techniques in recyclable waste paper sorting. In recycling, waste papers are segregated into various grades as they are subjected to different recycling processes. Highly sorted paper streams facilitate high quality end products and save processing chemicals and energy. From 1932 to 2009, different mechanical and optical paper sorting methods have been developed to fill the paper sorting demand. Still, in many countries including Malaysia, waste papers are sorted into different grades using a manual sorting system. Because of inadequate throughput and some major drawbacks of mechanical paper sorting systems, the popularity of optical paper sorting systems has increased. Automated paper sorting systems offer significant advantages over human inspection in terms of worker fatigue, throughput, speed, and accuracy. This research attempts to develop a smart vision sensing system that is able to separate the different grades of paper using first-order features. To construct a template database, a statistical approach with intra-class and inter-class variation techniques are applied to the feature selection process. Finally, the K-nearest neighbor (KNN) algorithm is applied for paper object grade identification. The remarkable achievement obtained with the method is the accurate identification and dynamic sorting of all grades of papers using simple image processing techniques.

Accurate characterization is an important issue in paper currency recognition system. This paper proposes a robust paper currency recognition method based on Hidden Markov Model (HMM). By employing HMM, the texture characteristics of paper currencies are modeled as a random process. The proposed algorithm can be used for distinguishing paper currency from different countries. A similarity measure has been used for the classification in the proposed algorithm. To evaluate the performance of the proposed algorithm, experiments have been conducted on more than 100 denominations from different countries. The results indicate 98% accuracy for recognition of paper currency.

This paper presents a fuzzy neural network predictive control scheme for studying the coagulation process of wastewater treatment in a paper mill. An adaptive fuzzy neural network is employed to model the nonlinear relationships between the removal rate of pollutants and the chemical dosages, in order to adapt the system to a variety of operating conditions and acquire a more flexible learning ability. The system includes a fuzzy neural network emulator of the reaction process, a fuzzy neural network controller, and an optimization procedure based on a performance function that is used to identify desired control inputs. The gradient descent algorithm method is used to realize the optimization procedure. The results indicate that reasonable forecasting and control performances have been achieved through the developed system.

This paper is concerned with a technique for detecting and monitoring abnormal paper formation variations in machine direction online in various frequency regions. A paper web is illuminated by two red diode lasers and the reflected light recorded as two time series of high resolution measurements constitute the input signal to the papermaking process monitoring system. The time series are divided into blocks and each block is analyzed separately. The task is treated as kernel based novelty detection applied to a multi-resolution time series representation obtained from the band-pass filtering of the Fourier power spectrum of the time series block. The frequency content of each frequency region is characterized by a feature vector, which is transformed using the kernel canonical correlation analysis and then categorized into the inlier or outlier class by the novelty detector. The ratio of outlying data points, significantly exceeding the predetermined value, indicates abnormalities in the paper formation. The experimental investigations performed have shown good repetitiveness and stability of the paper formation abnormalities detection results. The tools developed are used for online paper formation monitoring in a paper mill.

The system described here is concerned with the revision of inductive learning theory, i.e. the induced concept representation is updated and extended by taking into account new observations, as yet unlearned by the system. One of the aims of this paper is to show that it is feasible to classify observations, which have a great number of unknown values (incomplete information), as well as to estimate those values by exploiting pertinently known values. Imprecise information is also envisaged. The system works in two stages. Stage One applies a non-incremental unsupervised induction method, which is a learning cycle, make up of three steps: (i) verification of input information; (ii) classification of observations; (iii) organization of the information into hierarchies of clusters, according to concept formation. Stage Two is an incremental unsupervised induction method, which classifies new observations allowing for the later inclusion of currently unknown values. In both stages, the system explores the elements of each cluster in order to discover their relationship. Another interesting point in our proposal is that its involved algorithms have polynomial complexity in order to deal with large databases. The whole mechanism is illustrated throughout this paper by making reference to experimental results obtained in the Periodic Table of Elements, PTE.

As printing machines attain higher operating speeds, paper feeding mechanisms are more prone to problems such as vibration, noise, and paper jams. To design reliable paper feeding mechanisms, designers first use heuristic knowledge to decide the basic specifications of the printer according to the user's requirements. In the preliminary design stage, the design process is represented hierarchically and the characteristics of each part are determined. Conceptually, the process is analogous to the methodology used in configuration design. In this research, the design process has been analyzed systematically by the OMT (object modeling technique) method, which integrates the object, functional, and dynamic models. The physical parts of the paper feeding mechanism are represented as objects, and the design knowledge and design constraints are represented by rules and methods. A commercial expert system shell, Nexpert Object, is adapted for the implementation. The inferred results are converted to create the BOM (bill of materials) data and are saved in a database for the detailed design and production stage.

Feature selection determines relevant features in the data. It is often applied in pattern classification, data mining, as well as machine learning. A special concern for feature selection nowadays is that the size of a database is normally very large, both vertically and horizontally. In addition, feature sets may grow as the data collection process continues. Effective solutions are needed to accommodate the practical demands. This paper concentrates on three issues: large number of features, large data size, and expanding feature set. For the first issue, we suggest a probabilistic algorithm to select features. For the second issue, we present a scalable probabilistic algorithm that expedites feature selection further and can scale up without sacrificing the quality of selected features. For the third issue, we propose an incremental algorithm that adapts to the newly extended feature set and captures `concept drifts' by removing features from previously selected and newly added ones. We expect that research on scalable feature selection will be extended to distributed and parallel computing and have impact on applications of data mining and machine learning.

In this paper, we present a novel deep learning-based architecture, which is under the scope of expert and intelligent systems, to perform accurate real-time tridimensional hand pose estimation using a single RGB frame as an input, so there is no need to use multiple cameras or points of view, or RGB-D devices. The proposed pipeline is composed of two convolutional neural network architectures. The first one is in charge of detecting the hand in the image. The second one is able to accurately infer the tridimensional position of the joints retrieving, thus, the full hand pose. To do this, we captured our own large-scale dataset composed of images of hands and the corresponding 3D joints annotations. The proposal achieved a 3D hand pose mean error of below 5 mm on both the proposed dataset and Stereo Hand Pose Tracking Benchmark, which is a public dataset. Our method also outperforms the state-of-the-art methods. We also demonstrate in this paper the application of the proposal to perform a robotic hand teleoperation with high success. This work has been supported by the Spanish Government TIN2016-76515R Grant, supported with Feder funds. This work has also been supported by a Spanish grant for PhD studies ACIF/2017/243

Acquiring knowledge for conversation modeling is an important task in the process of building a Conversational Agent (Chatbot). However, it is a quite difficult process that requires too much time and efforts. To overcome these difficulties, in this paper, we demonstrate a novel methodology for the automatic conversational knowledge extraction from an existing Chatbot. Extracted knowledge will be used as training dataset for building a Neural Network Conversational Agent. The experiments in the paper show that our proposed approach can be used for the automatic knowledge extraction from any type of Chatbot on the Internet. The experiment that is presented in this paper has two phases. In the first phase, we present a methodology for the conversational knowledge extraction. In the second phase of the experiment, we introduce a methodology for building a new Neural Conversational Agent using a deep learning Neural Network framework. The key novelty of our proposed approach is the automated machine-machine conversational knowledge sharing and reuse. This is an important step towards building the new conversational agents skipping the difficult and time-consuming procedure of knowledge acquisition.

This paper presents a new image processing and analysis technique for the quality evaluation of cell nuclei to support medical diagnostics in breast cancer. The technique allows cell nuclei that are deformed or overlapped by biological material to be reconstructed. The paper proposes a sensitivity and similarity approach, enriching the PatchMatch correspondence algorithm in accurate cell reconstruction. Its application in reconstruction processes enables accelerated computations and an increased probability of obtaining appropriate segmentation results. The numerical results demonstrate that the developed system allows for automatic and effective cell nuclei reconstruction with an acceptable average area accuracy level above 85% compared with manual human results (assuming manual segmentation as a true value). The reconstruction system allows for the recovery of the proper shape of the analyzed distorted cells very rapidly and in a repeatable manner. An additional advantage of the procedure is that the nuclei area overlapped by artifacts or other cells can be determined. The experimental results prove the high utility of the method in final HER2 gene amplification assessment in breast cancer images.

Value efficiency analysis is an approach to incorporate the Decision-Maker's (DM) priori knowledge (preference information) into efficiency analysis. This paper presents a multi-objective programming model to measure the value efficiency. The model's feasible region is the production possibility set in data envelopment analysis (DEA) and the objective functions are the value efficiency scores corresponding to input and output components. Considering the structure of the proposed model and the concept of value efficiency, the step method (STEM) method to solve the model and gain the value efficiency is used in this paper. By selecting MPU (most preferred unit) units based on the manager's choice, the suitable benchmarks and value efficiency scores could be obtained corresponding to the unit under evaluation, then one could see that the obtained benchmarks and value efficiency scores are strongly dependent on the selected MPU units. Finally, a numerical example is proposed to illustrate the credit of the proposed approach. The method has been applied for real world data.

Many clustering algorithms categorized as K-clustering algorithm require the user to predict the number of clusters (K) to do clustering. Due to lack of domain knowledge an accurate value of K is difficult to predict. The problem becomes critical when the dimensionality of data points is large; clusters differ widely in shape, size, and density; and when clusters are overlapping in nature. Determining the suitable K is an optimization problem. Automatic clustering algorithms can discover the optimal K. This paper presents an automatic clustering algorithm which is superior to K-clustering algorithm as it can discover an optimal value of K. Iterative hill-climbing algorithms like K-Means work on a single solution and converge to a local optimum solution. Here, Genetic Algorithms (GAs) find out near global optimum solutions, i.e. optimal K as well as the optimal cluster centroids. Single-objective clustering algorithms are adequate for efficiently grouping linearly separable clusters. For non-linearly separable clusters they are not so good. So for grouping non-linearly separable clusters, we apply Multi-Objective Genetic Algorithm (MOGA) by minimizing the intra-cluster distance and maximizing inter-cluster distance. Many existing MOGA based clustering algorithms are suitable for either numeric or categorical features. This paper pioneered employing MOGA for automatic clustering with mixed types of features. Statistical testing on experimental results on real-life benchmark data sets from the University of California at Irvine (UCI) machine learning repository proves the superiority of the proposed algorithm.

Text summarization has become an important research area, especially in the biomedical domain, where information overload is a major problem. In this paper, we propose a novel biomedical text summarization system that combines two popular data mining techniques: clustering and frequent itemset mining. Biomedical paper is expressed as a set of biomedical concepts using the UMLS metathesaurus. The K-means algorithm is used to cluster similar sentences. Then, the Apriori algorithm is applied to discover the frequent itemsets among the clustered sentences. Finally, the salient sentences from each cluster are selected to build the summary using the discovered frequent itemsets. For the evaluation step, we selected randomly 100 biomedical papers from the BioMed Central database full-text, and we evaluated the performances of our system by comparing the resulting summaries with the abstracts of these papers using the ROUGE metrics in term of recall, precision, and F-measure. We also compared the obtained summaries with those achieved by five well-known summarizers: TextRank, TextTeaser, SweSum, ItemSet Based Summarizer, Microsoft AutoSummarize, and two baselines: summarization using only the frequent itemsets mining (FRQ-CL), and summarization using only the clustering (CL-FRQ). The results demonstrate that this combination can successfully enhance the summarization performances, and the proposed system outperforms other tested summarizers.

This research introduces the Recursive General Regression Neural Network Oracle (R-GRNN Oracle) and is demonstrated on several binary classification datasets. The traditional GRNN Oracle classifier (Masters et al., 1998) combines the predictive powers of several machine learning classifiers by weighing the amount of error each classifier has on the final predictions. Each classifier is assigned a weight based on the percentage of errors it contributes to the final predictions as the classifiers evaluate the dataset. The proposed R-GRNN Oracle is an enhancement to the GRNN Oracle in which the proposed algorithm consists of an oracle within an oracle – where the inner oracle acts as a classifier with its own predictions and error contribution. By combining the inner oracle with other classifiers, the R-GRNN Oracle produces superior results. The classifiers considered in this study are: Support Vector Machine (SVM), Multilayer Perceptron (MLP), Probabilistic Neural Network (PNN), Gaussian Naive Bayes (GNB), K-Nearest Neighbor (KNN), and Random Forest (RF). To demonstrate the effectiveness of the proposed approach, several datasets were used, with the primary one being the publicly available Spambase dataset. The predictions of SVM, MLP, KNN, and RF were used to create the first GRNN Oracle, which was then enhanced with the high performances of SVM and RF to create the second oracle, the R-GRNN Oracle. The combined recursive model was 93.24% accurate using 10-fold cross validation, higher than the 91.94% of the inner GRNN Oracle and the 91.29% achieved by RF, the highest performance by a stand-alone classifier. The R-GRNN Oracle was not only the most accurate, but it also had the highest AUC, sensitivity, specificity, precision, and F1-score (97.99%, 91.86%, 94.40%, 93.28%, and 92.57%, respectively). The research contribution of this paper is introducing the concept of recursion (a concept not fully explored in machine learning models and applications) and testing this structure's ability on further enhancing the performance of the traditional oracle. The recursive model has also been applied to several other datasets: The Human Resources, Bank Marketing, and Monoclonal Gammopathy of Undetermined Significance (MGUS) datasets. The results of these implementations are summarized in this paper.

This paper proposes a method of recognizing the intention of human activity based on percept sequences that represent the activities of daily living (ADL) in a residential space. Based on the activity intention ontology representing actions, poses, and nearby objects related to human activity intentions, the proposed method identifies a human activity intention by using the event calculus when a percept sequence is entered. It is very difficult to recognize ADL occurring in various places in a residence by using regular percept sequences without error. Furthermore, a human activity can have complex intentions. To solve these problems, this paper proposes an activity intention recognition process consisting of three steps. First, the activity intention inference step recognizes the intention of a given percept sequence based on the activity intention ontology. Second, the complex intention-identifying step determines whether to terminate the previously maintained activity intention or continue maintaining it in a complex manner based on the newly inferred activity intention. Third, the uncertainty handling step corrects an inaccurate activity intention caused by an error in the percept sequence. To evaluate the proposed method presented in this paper, activity intention recognition experiments were conducted based on the ADL data that were collected for six households of elderly persons older than 70 years. The results showed that the proposed method has a precision of 99.62% and recall of 92.83%.

This paper presents a new semantic kernel for classification of high-dimensional data in the framework of Support Vector Machines (SVM). SVMs have gained widespread application due to their relatively higher accuracy. The efficacy of SVMs, however, depends upon the separation of the data itself as well as the kernel function. Text data, for instance, is difficult to classify due to synonymy and polysemy in its contents, having multi-topical instances that can result in mislabeling, and being highly sparse in the bag-of-words representation. While the soft margin parameter and kernel tricks are used in SVM to deal with outliers and non-linearly separable data, using data statistics and correlation has not been fully explored in the literature. This paper explore the use co-similarity (i.e., soft co-clustering) to find latent relationships between documents motivated by the success of co-clustering and subspace clustering methods. It has been shown that the use of weighted higher-order paths between instances in the data can be a good measure of similarity values which can then be used for both classification and to correct mislabeled (or outlier) data in the training set. The proposed kernel is generic in nature and suitable for sparse, dyadic data where direct co-occurrences are not necessary common as in the case of textual data, link-analysis in social media networks, co-authorship, etc. It also studies the impact of noise in the training data and provides a technique to re-label such instances. It is also observed that re-labelling of selected training data reduces the adverse effect of outliers or label noise and can greatly improve the classification of the test data. To the best of our knowledge, we are the first to introduce a supervised co-similarity based kernel function and also provide mathematical formulation to show that it is a valid Mercer's kernel. Our experiments show that the proposed framework outperforms current and state-of-the-art methods in terms of classification accuracy and is more resilient to label noise.

Software clone detection is an emerging research area in the field of software engineering. Software systems are subjected to continuous modifications in source code to improve the performance of the software, which may lead to code redundancy. Duplicate code/code clone is a piece of code reworked several times in software programs due to copy paste activity or reusability of existing software. Code clone is a prime subject in software evolution. Detection of software clones at the time of software evolution may improve the performance of software and reduce the maintenance cost and effort. This paper proposes metric based methods to detect code clones, as software clone is a universal problem in large scale programming environment. This paper introduces two metric based approaches to detect code clones by comparing (i) Control Statement Features (ii) Program Features like different types of statements, operators and operands. In order to demonstrate the effectiveness of the proposed approaches, extensive experiments are conducted on two datasets, C projects of Bellon's benchmark dataset and student lab programs (SLP).The methods efficiently identify similar functional clones. Proposed models only find similarity of whole programs but intelligent enough to highlight similar code segments across program files.

Low quality of business process event logs, as determined by anomalous and missing values, is often unavoidable in practical contexts. The output of process analysis that uses event logs with missing and anomalous values is also likely to be of low quality, thus decreasing the quality of any decisions based on it. While previous work has focused on reconstructing missing events in an event log or removing anomalous traces, in this paper we focus on detecting anomalous values and reconstructing missing values at the level of attributes in event logs. We propose methods based on autoencoders, which are a class of neural networks that can reconstruct their own input and are particularly suitable to learn a model of the complex relationships among attribute values in an event log. These methods do not rely on any a-priori knowledge about the business process that generated an event log and are evaluated using real world and artificially-generated event logs. The paper also discusses a qualitative analysis of the impact of event log cleaning and reconstruction on the output of process discovery. The proposed approach shows remarkable performance regarding activity labels and timestamps in artificial event logs. The performance in the case of real world event logs, in particular timestamp anomaly detection, is lower, which may be due to high variability of attribute values in the chosen event logs. Process models discovered from reconstructed event logs are characterised by lower variability of allowed behaviour and, therefore, are more usable in practice.

Classification of text based on its substance is an essential part of analysis to organize enormously large text data and to mine the salient information contained in it. It is gaining greater attention with the surge in the volume of on-line data available. Classical algorithms like k-NN (k-nearest neighbor), SVM (Support Vector Machine) and their variations have been observed to yield only reasonable results in addressing the problem, leaving enough room for further improvement. A class of algorithms commonly referred to as Sparse Methods has been emerged recently from compressive sensing and found numerous effective applications in many areas of data analysis and image processing. Sparse Methods as a tool for text analysis is an alley that is largely unexplored rigorously. This paper presents exploration of sparse representation-based methods for text classification. Based on the success of sparse representation based methods in different areas of data analysis, we intuitively hypothesized that it should work well on text classification problems as well. This paper empirically reinforces the hypothesis by testing the method on Reuters and WebKB data sets. The empirical results on Reuters and WebKB benchmark data show that it can outperform classical classification algorithms like SVM and k-NN. It has been observed that obtaining the basis of representation and sparse codes are computationally costly operations affecting the performance of the system. We also propose a class-wise dictionary refinement algorithm and dynamic dictionary selection algorithm to make sparse coding faster. The addition of dictionary refinement to the classification system not only reduces the time taken for sparse coding but also gives improved classification accuracy. The outcomes of the study are empirical verification of sparse representation classifier as a text classification tool and a computationally efficient solution for the bottleneck operation of sparse coding.

Facial image manipulation is a particular instance of digital image tampering, which is done by compositing a region from one facial image into another facial image. Fake images generated by facial image manipulation now spread like wildfire on news websites and social networks, and are considered the greatest threat to press freedom. Previous research relied heavily on handcrafted features to analyze tampered regions which were inefficient and time-consuming. This paper introduces a framework that accurately detects manipulated face image using deep learning approach. The original contributions of this paper include (1) a customized convolutional neural network model for Manipulated Face (MANFA) identification; it contains several convolutional layers that effectively extract features of multi-levels of abstraction from a tampered region. (2) A hybrid framework (HF-MANFA) that uses Adaptive Boosting (AdaBoost) and eXtreme Gradient Boosting (XGBoost) to deal with the imbalanced dataset challenge. (3) A large manipulated face dataset that is manually collected and validated. The results from various experiments proved that proposed models outperformed existing expert and intelligent systems which were usually used for the manipulated face image detection task in terms of area under the curve (AUC), computational complexity, and robustness against imbalanced datasets. As a result, the presented framework will motivate the finding of a more powerful altered face images detection method and encourages the integration of the proposed model in applications that have to deal with manipulated images regularly.

Predictive analytics is an essential capability in business process management to forecast future status and performance of business processes. In this paper, we focus on one particular predictive monitoring task that is solved using classification techniques, i.e. predicting the next event in a case. Several different classifiers have been recently employed in the literature in this task. However, a quantitative benchmark of different classifiers is currently lacking. In this paper, we build such a benchmark by taking into account 20 classifiers from five families, i.e. trees, Bayesian, rule-based, neural and meta classifiers. We employ six real-world process event logs and consider two different sampling approaches, i.e. case and event-based sampling, and three different validation methods in order to acquire a comprehensive evaluation about the classifiers’ performance. According to our benchmark, the classifier most likely to be the overall superior performer is the credal decision tree (C-DT), followed by the other top-4 performers, i.e. random forest, decision tree, dagging ensemble, and nested dichotomies ensemble. We also provide a qualitative discussion of how features of an event log can affect the choice of best classifier.

Machine learning has become a pillar of today’s expert and intelligent systems where a special attention has been drawn to unsupervised methods. Support vector data description is one of the most interesting methods proposed in this context that addresses two main challenges: the absence of ground truth labels and the presence of outliers in data. It is well known that the ultimate problem has a great impact on a system performance since it typically prevents and limits its generalization and hence its failure. Even though the strengths of the Support vector data description method, its application into multi class and large scale context, as computer vision systems, raises bottleneck issues due to its runtime complexity and mono-class context application. We propose, in this paper, an approach inspired from the movement behavior of particle swarm and relies on the use of core set paradigm to overcome the aforementioned drawbacks of Support vector data description method. Indeed, since the proposed method uses some parameters corresponding to kernel function and swarm’s movement formulation, we present and study in this paper some strategies to ensure a dynamic adaptation of these parameters. The main qualities of the proposed method is that it provides a fast scheme for clustering a set of points without neither a prior knowledge on the number of naturally occurring cluster in the data nor an assumption on clusters shapes. A quantitative and comparative performance assessment is carried out over the Berkeley and BuffaloXiph datasets. The results witness the efficiency of the method in terms of both classification accuracy and computational performance. These quantitative assessments reinforce the significance as well as the importance of embedding the proposed method in other intelligent systems application areas.

Extensive research has shown that collaboration has substantial impact on supply chains performance. However, most of previous research has focused on information sharing-based collaboration models that require considerable effort to be implemented, such as information exchange supply chain (IESC). This paper introduces a new efficient collaboration model (IS-OUT) for multi-echelon supply chains. The IS-OUT model relies upon the ordering mechanism of the classical order-up-to policy (OUT). In traditional supply chains (TSC), although OUT replenishment orders include two pieces of information: demand forecast and inventory position balance, they are transferred to upstream echelons as single-quantity orders. In IS-OUT, the order information is transferred to the upstream echelons as two component parts of information to provide better coordination in supply chains. In this paper, the mathematical formulation of IS-OUT is presented, and simulation is adopted to compare the performance of TSC, IS-OUT, and IESC under special conditions and assumptions, considering various performance metrics. Although the results indicate that IESC model offers higher performance mostly than IS-OUT, IS-OUT outperforms or at least is comparable with IESC under a few conditions. Since IS-OUT model is easier to implement than IESC, IS-OUT provides a compromise between extent and information sharing requirements, and performance efficiency. The results should help a decision maker to select the model that optimizes its operating environment.

Smart cruise control (SCC) is one of the representative systems of advanced driver assistance systems (ADAS). Conventional SCC systems have a problem in that they cannot detect a cut-in vehicle in a short distance because the field of view (FOV) of their sensors for detecting vehicles ahead is limited. To solve this problem, this paper proposes a novel method using surround view monitoring (SVM) cameras which can observe the surroundings of a host vehicle without blind spots. To reduce the variation of projected appearance according to the relative position, this paper proposes to exploit multiple virtual camera images (rotational images). In each rotational image, vehicles are detected using a part-based method: tires are detected and then vehicles are detected by combining the tires. Detected tires in multiple rotational images are integrated and tracked in 3D space. Finally, the proposed method determines whether a vehicle has intruded into the cut-in zone based on the position and orientation of the vehicle. The proposed method succeeds to detect vehicle even by a simple method thanks to the rotational image even without any special H/W. The proposed method uses only the sensor adopted in mass-produced vehicles and it is very practical. Moreover, the method can estimate the surrounding vehicle position accurately enough to be used in the braking control system. The performance of the proposed method is evaluated with various driving situations, of which the total length is 447.8 min (43.5 min for the test track and 404.3 min for natural driving). Among 211 cut-in events, 208 events are correctly warned and only 3 false warnings occur. Although the proposed method simultaneously processes the rotational images generated from three SVM cameras (forward, left, and right), the average processing time per frame takes about 58 ms.

The present paper aims to show the necessity of information augmentation to cope with the natural decrease in information content in multi-layered neural networks. It is natural to make an effort to collect as much information as possible, because it is impossible to know which information is necessary or useful before learning. Thus, the present paper tries to force neural networks to form new network configurations to have as much information as possible, contrary to the conventional approach of information reduction, such as many types of regularization. For information augmentation, we use the self-organizing map (SOM), which can over-represent inputs and produce as many similar weights as possible. The method was applied to two data sets: the banknote authentication data set and the character recognition data set. In both experimental results, it was confirmed that redundant and excessive information generation in terms of the excessive number of connection weights was connected with improved generalization.

One of the concepts that has gained important attention from economic and management-science research is that of the concept of returns to scale (RTS). There are several research papers on the behavior of the right and left RTS, as two specific directions of RTS, based on Data Envelopment Analysis (DEA) models. However, the main weakness of most of these methods is that they are based on the defined parameters, which itself leads to a high sensitivity of the models to variations in the magnitudes of the parameters and, thus, unreliable results. In this paper, we specially address this issue and propose a simple procedure for detecting the right and left RTS classification with an important feature that is independent of any predetermined parameters. In addition to the type of RTS, we also suggest a method to determine the value of the right and left RTS corresponding to each of the efficient DMUs. Finally, the superiority of the proposed parameter-free procedure over a parameter-based methods is elaborated by a real case in the banking system.

It is a very challenging and important task to adaptively adjust the scale factor F and the crossover rate Cr for Differential Evolutionary (DE) algorithms. Most recent adaptive techniques were designed to generate parameters randomly based on successful trial values during the previous evolving process, lacking explicit guidelines to generate appropriate values. This paper proposes a novel parameter adaption strategy, which could incorporate promising F and Cr pairs extracted by using Association Rule Mining (ARM) into DE algorithms. First, all successful F and Cr values generated by their original methods are recorded during the whole evolution, resulting in an increasing dataset. Second, we discretize the dataset and extract the most frequent itemset of parameters by using a modified version of the widely used Apriori algorithm. Third, a greedy operator is developed to generate new parameters in the next generation by comparing the presented ARM-based and original-method-based fitness values. The presented technique provides an additional pair of F and Cr values to be evaluated, without replacing existing strategies for the control parameters. The main contribution of this paper is that we propose a novel way, which utilizes information generated during the evolutionary process, to enhance exploration capabilities by adjusting control parameters. Experimental results demonstrate that the proposed ARM-based parameter adaptive strategy is able to enhance performances of some state-of-the-art DE variants. Further, this methodology might be helpful for other control parameters of Evolutionary Algorithms (EA).

Clustering of multi-view data has received increasing attention since it explores multiple views of data sets aiming at improving clustering accuracy. Particle Swarm Optimization (PSO) is a well-known population-based meta-heuristic successfully used in cluster analysis. This paper introduces two hybrid clustering methods for multi-view relational data. These hybrid methods combine PSO and hard clustering algorithms based on multiple dissimilarity matrices. These methods take advantage of the global convergence ability of PSO and the local exploitation of hard clustering algorithms in the position update step, aiming to improve the balance between exploitation and exploration processes. Moreover, the paper provides adapted versions of 11 fitness functions suitable for vector data aiming at dealing with multi-view relational data. Two performance criteria were used to evaluate the clustering quality using the two proposed methods over eleven real-world data sets including image and document data sets. Among new findings, it was observed that the top three fitness functions are Silhouette index, Xu index and Intra-cluster homogeneity. The performance of the proposed algorithms was compared with previous single and multi-view relational clustering algorithms. The results show that the proposed methods significantly outperformed the other algorithms in the majority of cases. The results reinforce the importance of the application of techniques such as PSO-based clustering algorithms in the field of expert systems and machine learning. Such application enhances classification accuracy and cluster compactness. Besides, the proposed algorithms can be useful tools in content-based image retrieval systems, providing good categorizations and automatically learning relevance weights for each cluster of images and sets of views.

Pen and paper-based value stream mapping (VSM) is the established tool for recording processes, identifying waste and deriving recommendations for action. However, today, its application in manufacturing industry requires a high level of effort and is challenging due to product and process complexity, as well as dynamics. To overcome these shortcomings, we are developing a methodology to apply process mining (PM) to internal logistics for a mixed-model assembly line. The methodology combines multidimensional process mining (MDPM) techniques with proven principles of lean production and VSM. Firstly, internal logistics is modelled using existing event data by automatically mapping physical logistics activities (e.g. transport, store). Secondly, to enable PM, the event data is transformed into enriched event logs. Thirdly, the MDPM approach contains (1) a discovery analysis, (2) a performance analysis and (3) a conformance analysis including a reference process classification for each individual part and process. Finally, a waste analysis and strategy for practitioners is designed to identify and prioritise wasteful parts and processes. The methodology has been applied and evaluated in a case study at a German automotive manufacturer. In the case study, we analysed 7500 parts and 15 reference processes. An analysis for each individual part and process has not been available yet. We could both identify part-specific root causes (e.g. a long lead time) and process-specific root causes (e.g. a low trace fitness). The main contribution of this paper is to provide an MDPM methodology for practitioners to enable a continuous recording, evaluation and waste analysis of each individual part and process within internal logistics.

The Rapidly-exploring Random Tree (RRT) algorithm is a popular algorithm in motion planning problems. The optimal RRT (RRT*) is an extended algorithm of RRT, which provides asymptotic optimality. This paper proposes Quick-RRT* (Q-RRT*), a modified RRT* algorithm that generates a better initial solution and converges to the optimal faster than RRT*. Q-RRT* enlarges the set of possible parent vertices by considering not only a set of vertices contained in a hypersphere, as in RRT*, but also their ancestry up to a user-defined parameter, thus, resulting in paths with less cost than those of RRT*. It also applies a similar technique to the rewiring procedure resulting in acceleration of the tendency that near vertices share common parents. Since the algorithm proposed in this paper is a tree extending algorithm, it can be combined with other sampling strategies and graph-pruning algorithms. The effectiveness of Q-RRT* is demonstrated by comparing the algorithm with existing algorithms through numerical simulations. It is also verified that the performance can be further enhanced when combined with other sampling strategies.

We introduce a multidimensional multiblock clustering (MDMBC) algorithm in this paper. MDMBC can generate overlapping clusters with similar values along clusters of dimensions. The parsimonious binary vector representation of multidimensional clusters lends itself to the application of efficient meta-heuristic optimization algorithms. In this paper, a hill-climbing (HC) greedy search algorithm has been presented that can be extended by several stochastic and population-based meta-heuristic frameworks. The benefits of the algorithm are demonstrated in a bi-clustering benchmark problem and in the analysis of the Leiden higher education ranking system, which measures the scientific performance of 903 institutions along four dimensions of 20 indicators representing publication output and collaboration in different scientific fields and time periods.

Every music composition has a composer at the core of its building block, molding it into a style of their own. The creative compositional style of a composer varies dynamically with every composer which is perishable and inimitable but cannot be preserved. This paper proposes a hybrid style transfer model which is an end-to-end approach to transfer style, compose as well as evaluate the style transfer accuracy in the domain of the target music composer. The paper focuses on 3 composer maestros Liszt, Chopin and Schubert taken from the Maestro dataset. The main step of music composer style transfer is accomplished in 3 systematic steps, training composer classifiers according to feature sets, generating a model of a composition in a particular composer’s style from noise, and finally style transfer and its evaluation. The 3 steps are interconnected in a way that each step is the building block for the next step. The GAN architecture of the style transfer step is built out of the GAN architecture of the second step. The compositions generated from each architecture is finally evaluated by the common pre-trained classifier of the first step. The highest accuracy obtained is 80% for composer classification using the maestro dataset, 77.27% for the classification of the generated style transferred version of a composition into the target composer class using the pre-trained classifiers.

This paper presents the Storybox Methodology which combines a novel framework for structuring knowledge and conversations around a story (D-PAF), with a live chatroom-based training approach that builds the conversation knowledge base via live chatroom interactions. Chatbots have achieved success as intelligent interfaces in education, health, sales and support, but their move towards mainstream adoption has been hindered by the large amount of development resources required, in terms of data collection, preparation, user testing and technical knowledge. The complexity of the development task often necessitates both a system author and a domain expert working effectively together, adding further complexity and risk. Overcoming these barriers could increase feasibility of chatbots in a range of expert contexts. In education, there are groups of learners who do not enjoy reading and writing. Storytelling chatbots might be able to introduce these groups to enjoyable new ways to read and write, having a beneficial impact on their education and future prospects. This paper proposes the Storybox Methodology for the rapid development of storytelling chatbots. Storybox is evaluated by creating, training and testing ‘The Ghost’, a chatbot enacting Hamlet's Ghost character from William Shakespeare's dramatic tragedy. The results showed that after a period of live chatbot training of only 25 training conversations, The Ghost was able to conduct convincing conversations with participants.

Lithology identification is an important part of petroleum drilling engineering. Accurate identification of lithology is the foundation to ensure the smooth operation of drilling engineering. Conventional lithology recognition mainly relies on human experience. The recognition accuracy depends on the level of technical personnel and the recognition response time is lagging. It is difficult to meet the demand. How to achieve rapid and intelligent recognition of lithology is one of the core technical problems faced by oil drilling. In order to solve this problem, this paper uses the mechanical specific energy and other ground parameters to establish a rapid and intelligent recognition method of lithology based on the simulated annealing optimization support vector machine model. In order to improve the accuracy of the model recognition, a large number of methods have been developed from two classes and multiple classes, simulation analysis results show that the lithology recognition model based on the principle of mechanical specific energy and the simulated annealing optimization support vector machine algorithm can predict a priori unknown data with a prediction accuracy of over 90%. Compared with the support vector machine model and the K-means model, simulated annealing optimization support vector machine is used for comparative analysis, the algorithm to establish a lithology recognition model has better performance and higher accuracy. The intelligent identification model of lithology based on the principle of mechanical specific energy and simulated annealing optimization support vector machine algorithm established in this paper can quickly and accurately identify lithology, and provide new technical support for oil drilling formation analysis.

With the rapid development of e-commerce, online reviews have played an increasingly important role in consumers' shopping intentions and behaviors. Therefore, how to effectively identify fake reviews has become one of the important issues that need to be resolved. Since the existing methods do not fully consider the time burst characteristics of reviews, this paper proposes a suspicion degree determining method based on the three-dimensional time series. Besides, combining the suspicion degree feature, review text features, and reviewer's behavior features together, this paper proposes a more comprehensive fake review identification framework. The yelp and amazon public data sets are carried out to verify the effectiveness of the proposed method, and the experimental results show that the proposed method outperforms the most advanced methods.

Nowadays the use of visual navigation in orchard is rapidly growing, while current research mainly focuses on orchard row detection or obstacle detection separately. The economic, fast and reliable navigation vision system that completing orchard row following and obstacle avoidance simultaneously and the obstacle avoidance algorithm for vision system without complex multiple navigation systems combining are needed further research. In this paper, a trinocular vision system for orchard vehicle is set up and the obstacle avoidance algorithm based on coupling of geometric constraint and virtual force field is designed for its vision system. First of all, based on analyzing the different vision system characteristics and the detection demand of orchard obstacle avoidance the trinocular vision system for orchard vehicle is set up based on a wide-angle camera and binocular stereo vision system. Then, the image processing algorithm of orchard row and obstacle detection is designed. For orchard row detection, the trunk regions enhancement algorithm is designed based on grayscale morphology filtering the trunk regions prediction algorithm is designed based on optical flow method to improve the trunk regions detection speed and accuracy. For obstacle detection, the background equalization algorithm based on H-channel image characteristics and the interference weaken algorithm with G-channel image characteristics are designed. Based on the row and obstacle detection results the algorithm for obstacle avoidance in orchard based on coupling of geometric constraint and virtual force field is designed. In the obstacle avoidance process, the virtual force field method is used as the mainly control of the vehicle obstacle avoidance process, and the geometric constraint between visual model of trinocular vision system and obstacle position is coupling with it to fulfill the avoidance demand of the obstacle with special shape and other condition that influence the virtual force field calculation. The experimental results show that the trinocular vision system for orchard vehicle built in this paper has strong adaptability to the actual environment which is accurate and stable for orchard row detection and obstacle detection simultaneously. The average deviation is 4.76 cm and 7.05 cm at 0.5 m/s and 1.0 m/s respectively, the average deviation of obstacle distance detection in the Z-axis direction is 3.18 cm, and in the X-axis direction is 0.45 cm. And the obstacle avoidance algorithm designed in this paper is effective in the orchard which can meet the actual production requirements. The research results will lay a foundation for the development of intelligent equipment and unmanned management in orchard.

This paper presents an approach to design better line plans for realistic cases. The planning problem is modelled as a realistic instance of the Transit Network Design and Frequency Setting Problem (TNDFSP). It incorporates additional assumptions taken from practice such as discrete frequencies, a subset of allowed terminal nodes and circular lines. A bi-objective memetic algorithm minimizes the average travel time (ATT) of the passengers and the fleet size. The method to generate the TNDFSP instance from real data is described in detail. Moreover, a new metric is proposed to compare different transit networks. In order to illustrate the approach, it is applied to the area of Utrecht, The Netherlands. The results show that the current network is modelled accurately and that the algorithm successfully generates alternative bus line plans within reasonable CPU time. It returns a subset of non-dominated solutions from which a compromise solution can be selected for practice. There are solutions with the same fleet size as the current solution, but a 6% lower ATT, or solutions with the same ATT, but a fleet size which is 19% smaller. Compared to the current network, the algorithm finds that it is convenient to substantially reduce the number of lines and to leave a small portion of demand unsatisfied. This paper also presents extensive experiments to test the impact of different assumptions about demand, bus capacity and minimum frequency.

Regular Ship hull maintenance is an essential for sustainability. The maintenance work of ship hulls that involve human labor suffers from many shortcomings. Maintenance robots have been introduced for drydocks to eliminate these shortcomings. An energy-efficient Complete Coverage Path Planning (CCPP) is a crucial requirement from a ship hull maintenance robot. This paper proposes a novel energy-efficient CCPP method based on Glasius Bioinspired Neural Network (GBNN) for a ship hull inspection robot. The proposed method accounts for a comprehensive energy model for path planning. This energy model reflects the energy usage of a ship hull maintenance robot due to changes in direction, distance, and vertical position. Furthermore, the proposed method is effective for dynamic workspaces since it performs online path planning. These are the major contributions made to state of the art by the work proposed in this paper. The behavior and the performance of the proposed method have been compared against state of the art through simulations considering Hornbill, a multipurpose ship hull maintenance robot. The validation confirms the ability of the proposed in realizing a complete coverage of a given dynamic workspace. According to the statistical outcomes of the comparison, the performance of the proposed method significantly surpasses that of the state-of-the-art methods in terms of energy usage. Therefore, the proposed method contributes to the development of energy-efficient CCPP methods for a ship hull maintenance robot.

In Data Envelopment Analysis (DEA), a variety of approaches have been used in the context of single-stage and basic serial two-stage systems to attain fairness in the evaluation of decision-making units (DMUs). Little work, however, has been done to address this challenge in a generalised two-stage structure featuring additional inputs in the second stage and a proportion of first-stage outputs as final outputs. In this paper, we argue that in this context, fairness is enhanced by increasing measures related to the discriminatory power and the weighting scheme of the method. We describe a mechanism that gives prominence to a more contemporary concept of fairness, incorporating diversity and inclusion of minority opinions. These aspects have, to our knowledge, not yet received explicit attention in the methodological development of DEA. We propose a novel combination of an additive self-efficiency aggregation model, a minimax secondary goal model, and the CRiteria Importance Through Inter-criteria Correlation (CRITIC) method, in order to promote these aspects of fairness, and thus achieve a better degree of cooperation between the stages of a DMU and among DMUs. The additive aggregation model is chosen over the alternative multiplicative approach for a variety of reasons relating to the emphasis on the intermediate products exchanged and the simplification. The minimax model offers peer evaluation in which each DMU aims to evaluate the worst of the others in the best possible light. Application of the CRITIC method to DEA addresses the aggregation problem within the cross-efficiency concept. Practical applications of this approach could include supporting the determination of training needs in job rotation manufacturing, or evaluation of sustainable supply chains. The paper includes a description of a numerical experiment, illustrating the approach.

Summarization, is to reduce the size of the document while preserving the meaning, is one of the most researched areas among the Natural Language Processing (NLP) community. Summarization techniques, on the basis of whether the exact sentences are considered as they appear in the original text or new sentences are generated using natural language processing techniques, are categorized into extractive and abstractive techniques. Extractive summarization has been a very extensively researched topic and has reached to its maturity stage. Now the research has shifted towards the abstractive summarization. The complexities underlying with the natural language text makes abstractive summarization a difficult and a challenging task. This paper presents a comprehensive review of the various works performed in abstractive summarization field. For this purpose, we have selected the recent papers on this topic from Elsevier, ACM, IEEE, Springer, ACL Anthology, Cornell University Library and Google Scholar. The papers are categorized according to the type of abstractive technique used. The paper lists down the various challenges and discusses the future direction for research in this field. Along with these, we have identified the advantages and disadvantages of various methods used for abstractive summarization. We have also listed down the various tools which have been used or developed by researchers for abstractive summarization. The paper also discusses the evaluation techniques being used for assessing the abstractive summaries.

Construction is a knowledge-intensive industry, in which organizations are known for the delivery of products and services, relying on different types of knowledge. To manage knowledge effectively, various information technology (IT) models and systems have been developed for knowledge management (KM) in the construction industry over the years. As the next generation of IT application in construction, building information modeling (BIM) is increasingly used today to aid KM. Compared to generic IT tools, BIM has some distinctive features, such as parametric modeling, virtual visualization and centralized platform. However, how to apply these features of BIM to better serve KM has not yet been well summarized and analyzed. In this research, 115 papers on IT-based KM and 73 papers on BIM-supported KM are reviewed, based on which an integrated framework is developed to describe the current status and future directions for KM in IT-generic and BIM-specific contexts. It is followed by a conceptual model of BIM-supported KM which shows the possible KM factors and their relationships in the BIM environment. This research highlights the transformation from IT-based KM into BIM-supported KM. It contributes to the identification of remaining challenges to BIM-supported KM and the elaboration of BIM-supported KM in construction research and practice.

People with type 1 diabetes must control their blood glucose level through insulin infusion either with several daily injections or with an insulin pump. However, estimating the required insulin dose is not easy. Recommender systems, mainly based on Case-Based Reasoning (CBR), are being developed to provide recommendations to users. These systems are designed to keep the experiences or cases of the user in a case-base, which requires maintenance to keep system's response accurate and efficient. This paper proposes a case-base maintenance methodology that combines case-base redundancy reduction and attribute weight learning. Contrary to previous approaches designed for classification problems, the maintenance methodology presented in this paper deals with numerical recommendations. It can manage a potentially huge case-base due to the combinatorial derived from the number of attributes used to represent a case. The proposed approach has been tested using the UVA/PADOVA type 1 diabetes simulator and the results demonstrate that it can accomplish better levels of accuracy than other insulin recommender systems mentioned in the literature, when a large number of attributes is considered This project has received funding from the European Union Horizon 2020 research and innovation programme under grant agreement No. 689810, www.pepper.eu.com/, PEPPER, and the grant of the University of Girona 20162018 (MPCUdG2016). The work has been developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 20142016)

This paper presents a comparative analysis of the outcomes achieved when two widely applied methods for supplier selection—the ‘technique for order of preference by similarity to ideal solution’ (TOPSIS) and ‘data envelopment analysis’—are applied to the problem of identifying the most preferred sustainable suppliers. Both fuzzy DEA and fuzzy TOPSIS are applied to a common dataset of logistics service providers in Sweden. The results reveal that TOPSIS outperforms DEA in terms of both calculation complexity and sensitivity to changes in the number of suppliers. However, output rankings from the two models are found to be less than perfectly correlated. The paper concludes that utilizing both methods, as applied to just a small number of evaluation criteria and a relatively low level of detail in the data, produces a useful pooled shortlist of potential sustainable suppliers. This can then form the basis for a second stage application where either of the methods may be applied to a greater number of criteria that are specified to a higher level of detail. Even more critically, the results also have the potential to point to specific aspects for discussion when negotiating price and service quality commitments with potential sustainable suppliers.

Identification of tumor in medical images is an important task which leads to increased survival chances. In many cases, tumor diagnosis in medical images is very complicated since the pixel intensities between benign and malignant tissues may be very close. The identification problem is defined as the intensification of the mutual data between the image pixel intensities and the desired areas (tumors in this paper), depending on a limitation of the overall size of the desired area borders. This paper introduces a new approach for image segmentation in differentiating between benign and malignant tissues through a nonparametric approach. The proposed approach assumes that the probability densities correlated with pixel intensities in the image for each region are not already known. The image intensity is used instead of using the probability. A nonparametric density estimation is proposed and formulates the theoretical data optimization problem by applying curve evolution methods and deriving the correlated gradient flows. It uses level-set techniques to achieve the resulting evolution. The algorithm is applied to 100 sets of different real data in patient images with benign or/and malignant tissues. The experimental results show that the proposed algorithm proved effective in tumor identification with an acceptable accuracy of 93.1%, especially in-patient images diagnosed with carcinoma at an early stage.

Activation functions play a crucial role in discriminative capabilities of the deep neural networks. They are also one of the main reasons for revival of the neural networks. Although, recent activation functions provide solution to vanishing and exploding gradient problems, but they do not have sufficient capacity to model non-linear data. This paper proposes a novel activation function, which imparts neural networks with capability to model non-linear dependencies in data. Behavior of proposed activation function remains non-saturating even with non-linear structure. The activation function provides distinct activation behaviors to different data range segments. The paper presents two different variants of proposed function. The first, “linear sigmoidal activation” function is a fixed structure activation function with function coefficients defined at the start of model design. Whereas second, “adaptive linear sigmoidal activation” function is a trainable function which can adapt itself according to the complexity of the given data. Both of the proposed models are tested against the state of the art activation functions on benchmark datasets (CIFAR-10, MNIST, SVHN, FER-2013). The proposed activation function is able to outperforms every known activation functions in all the tests.

This paper presents an application of a recently well performing evolutionary algorithm for continuous numerical optimization, Success-History Based Adaptive Differential Evolution Algorithm (SHADE) including Linear population size reduction (L-SHADE), to an expert system for underwater glider path planning (UGPP). The proposed algorithm is compared to other similar algorithms and also to results from literature. The motivation of this work is to provide an alternative to the current glider mission control systems, that are based mostly on multidisciplinary human-expert teams from robotic and oceanographic areas. Initially configured as a decision-support expert system, the natural evolution of the tool is targeting higher autonomy levels. To assess the performance of the applied optimizers, the test functions for UGPP are utilized as defined in literature, which simulate real-life oceanic mission scenarios. Based on these test functions, in this paper, the performance of the proposed application of L-SHADE to UGPP is aggregated using statistical analyis. The depicted fitness convergence graphs, final obtained fitness plots, trajectories drawn, and per-scenario analysis show that the new proposed algorithm yields stable and competitive output trajectories. Over the set of benchmark missions, the newly obtained results with a configured L-SHADE outperforms existing literature results in UGPP and ranks best over the compared algorithms. Moreover, some additional previously applied algorithms have been reconfigured to yield improved performance. Thereby, this new application of evolutionary algorithms to UGPP contributes significantly to the capacity of the decision-makers, when they use the improved UGPP expert system yielding better trajectories.

Deep convolutional models have been able to get extraordinary results in visual, speech and textual processing domains. Nevertheless, the complexity of manifold of images in data space necessitates the deep networks to have a large number of parameters and subsequently these deep models prone to redundancy and overfitting. One of the most effective methods to tackle this problem is devising special regularization methods in the context of convolutional models. In this paper, after presenting and discussing the most important models in the field, sparse feature maps are proposed and employed as a penalty term in the cost function. This conducts the learning process to construct kernels so that the feature maps at every point are sparse along the depth. A sparse representation is able to adapt to the varying level of information of natural images and can help to extract more independent and informative representations. Also, the DropMaps concept is proposed and employed in the last convolutional layer of the model. This technique applies dropout on the feature maps that causes coincidence of feature maps to be avoided. As is shown in this paper, sparse feature maps and DropMaps can handle the problem of overfitting in large models for tiny images. We have studied the effect of the sparsity rate on the accuracy of the model, and it is observed the accuracy of the test dataset reaches its maximum at a sparsity rate of 0.05. Moreover, by designing appropriate learning rate curves, we were able to obtain ensemble machines with much less cost for training. It is noticeable that the test accuracy is higher than the validation accuracy of the ensemble, indicating that the model has not overfitted. In the input of the model, a random online preprocessing layer is employed for the training phase that helps regularization of the model. Comparing input space and feature space of the model we found that the proposed network is able to successfully separate images of different classes. Finally, testing the proposed model with MNIST dataset has shown that the test set can be classified with accuracy 99.75. The same test with CIFAR 10 dataset attained an accuracy of 94.05.