Search results

The paper aims to explore whether topic analysis (identification of the core contents, trends and topic distribution in the target field) can be performed using a more low-cost and easily applicable method that relies on a small dataset, and how we can obtain this small dataset based on the features of the publications.

The paper proposes a topic analysis method based on prolific and authoritative researchers (PARs). First, the authors identify PARs in a specific discipline by considering the number of publications and citations of authors. Based on the research publications of PARs (small dataset), the authors then construct a keyword co-occurrence network and perform a topic analysis. Finally, the authors compare the method with the traditional method.

The authors found that using a small dataset (only 6.47% of the complete dataset in our experiment) for topic analysis yields relatively high-quality and reliable results. The comparison analysis reveals that the proposed method is quite similar to the results of traditional large dataset analysis in terms of publication time distribution, research areas, core keywords and keyword network density.

Expert opinions are needed in determining the parameters of PARs identification algorithm. The proposed method may neglect the publications of junior researchers and its biases should be discussed.

This paper gives a practical way on how to implement disciplinary analysis based on a small dataset, and how to identify this dataset by proposing a PARs-based topic analysis method. The proposed method presents a useful view of the data based on PARs that can produce results comparable to traditional method, and thus will improve the effectiveness and cost of interdisciplinary topic analysis.

This paper proposes a PARs-based topic analysis method and verifies that topic analysis can be performed using a small dataset.

The study aims to provide a quick-and-robust multifactorial screening technique for early detection of statistically significant effects that could influence a product's life-time performance.

The proposed method takes advantage of saturated fractional factorial designs for organizing the lifetime dataset collection process. Small censored lifetime data are fitted to the Kaplan–Meier model. Low-percentile lifetime behavior that is derived from the fitted model is used to screen for strong effects. A robust surrogate profiler is employed to furnish the predictions.

The methodology is tested on a difficult published case study that involves the eleven-factor screening of an industrial-grade thermostat. The tested thermostat units are use-rate accelerated to expedite the information collection process. The solution that is provided by this new method suggests as many as two active effects at the first decile of the data which improves over a solution provided from more classical methods.

To benchmark the predicted solution with other competing approaches, the results showcase the critical first decile part of the dataset. Moreover, prediction capability is demonstrated for the use-rate acceleration condition.

The technique might be applicable to projects where the early reliability improvement is studied for complex industrial products.

The proposed methodology offers a range of features that aim to make the product reliability profiling process faster and more robust while managing to be less susceptible to assumptions often encountered in classical multi-parameter treatments.

In this paper, we present a study about an automated system for monitoring underwater ecosystems. The system here proposed is based on the fusion of different deep learning methods. We study how to create an ensemble based of different Convolutional Neural Network (CNN) models, fine-tuned on several datasets with the aim of exploiting their diversity. The aim of our study is to experiment the possibility of fine-tuning CNNs for underwater imagery analysis, the opportunity of using different datasets for pre-training models, the possibility to design an ensemble using the same architecture with small variations in the training procedure.

Our experiments, performed on 5 well-known datasets (3 plankton and 2 coral datasets) show that the combination of such different CNN models in a heterogeneous ensemble grants a substantial performance improvement with respect to other state-of-the-art approaches in all the tested problems. One of the main contributions of this work is a wide experimental evaluation of famous CNN architectures to report the performance of both the single CNN and the ensemble of CNNs in different problems. Moreover, we show how to create an ensemble which improves the performance of the best single model. The MATLAB source code is freely link provided in title page.

We propose an oversampling technique to increase the true positive rate (sensitivity) in classifying imbalanced datasets (i.e., those with a value for the target variable that occurs with a small frequency) and hence boost the overall performance measurements such as balanced accuracy, G-mean and area under the receiver operating characteristic (ROC) curve, AUC. This oversampling method is based on the idea of applying the Synthetic Minority Oversampling Technique (SMOTE) on only a selective portion of the dataset instead of the entire dataset. We demonstrate the effectiveness of our oversampling method with four real and simulated datasets generated from three models.

The purpose of this paper is to propose a new deep learning-based model to carry out better maintenance for naval propulsion system.

This model is constructed by integrating different deep learning algorithms. The basic idea is to change the connection structure of the deep neural network by introducing a residual module, to limit the prediction output to a reasonable range. Then, connect the Deep Residual Network (DRN) with a Generative Adversarial Network (GAN), which helps achieve data expansion during the training process to improve the accuracy of the assessment model.

Study results show that the proposed model achieves a better prediction effect on the dataset. The average performance and accuracy of the proposed model outperform the traditional models and the basic deep learning models tested in the paper.

The proposed model proved to be better performed naval propulsion system maintenance than the traditional models and the basic deep learning models. Therefore, our model may provide better maintenance advice for the naval propulsion system and will lead to a more reliable environment for offshore operations.

The purpose of this paper is to assign topic‐specific ratings to web pages.

The paper uses power iteration to assign topic‐specific rating values (called relevance) to web pages, creating a ranking or partial order among these pages for each topic. This approach depends on a set of pages that are initially assumed to be relevant for a specific topic; the spatial link structure of the web pages; and a net‐specific decay factor designated ξ.

The paper finds that this approach exhibits desirable properties such as fast convergence, stability and yields relevant answer sets. The first property will be shown using theoretical proofs, while the others are evaluated through stability experiments and assessments of real world data in comparison with already established algorithms.

In the assessment, all pages that a web spider was able to find in the Nordic countries were used. It is also important to note that entities that use domains outside the Nordic countries (e.g..com or.org) are not present in the paper's datasets even though they reside logically within one or more of the Nordic countries. This is quite a large dataset, but still small in comparison with the entire worldwide web. Moreover, the execution speed of some of the algorithms unfortunately prohibited the use of a large test dataset in the stability tests.

It is not only possible, but also reasonable, to perform ranking of web pages without using Markov chain approaches. This means that the work of generating answer sets for complex questions could (at least in theory) be divided into smaller parts that are later summed up to give the final answer.

This paper contributes to the research on internet search engines.

This study proposes a collaborative federated learning (CFL) framework to address personal data transmission and retention issues for artificial intelligence (AI)-enabled video surveillance in public spaces.

This study examines specific challenges for long-term people monitoring in public spaces and defines AI-enabled video surveillance requirements. Based on the requirements, this study proposes a CFL framework to gradually adapt AI models’ knowledge while reducing personal data transmission and retention. The framework uses three different federated learning strategies to rapidly learn from different new data sources while minimizing personal data transmission and retention to a central machine.

The findings confirm that the proposed CFL framework can help minimize the use of personal data without compromising the AI model's performance. The gradual learning strategies help develop AI-enabled video surveillance that continuously adapts for long-term deployment in public spaces.

This study makes two specific contributions to advance the development of AI-enabled video surveillance in public spaces. First, it examines specific challenges for long-term people monitoring in public spaces and defines AI-enabled video surveillance requirements. Second, it proposes a CFL framework to minimize data transmission and retention for AI-enabled video surveillance. The study provides comprehensive experimental results to evaluate the effectiveness of the proposed framework in the context of facial expression recognition (FER) which involves large-scale datasets.

The implementation of hierarchic agglomerative methods of cluster anlaysis for large datasets is very demanding of computational resources when implemented on conventional computers. The ICL Distributed Array Processor (DAP) allows many of the scanning and matching operations required in clustering to be carried out in parallel. Experiments are described using the single linkage and Ward's hierarchical agglomerative clustering methods on both real and simulated datasets. Clustering runs on the DAP are compared with the most efficient algorithms currently available implemented on an IBM 3083 BX. The DAP is found to be 2.9–7.9 times as fast as the IBM, the exact degree of speed‐up depending on the size of the dataset, the clustering method, and the serial clustering algorithm that is used. An analysis of the cycle times of the two machines is presented which suggests that further, very substantial speed‐ups could be obtained from array processors of this type if they were to be based on more powerful processing elements.

This study aims to differential diagnosis of some diseases using classification methods to support effective medical treatment. For this purpose, different classification methods based on data, experts’ knowledge and both are considered in some cases. Besides, feature reduction and some clustering methods are used to improve their performance.

First, the performances of classification methods are evaluated for differential diagnosis of different diseases. Then, experts' knowledge is utilized to modify the Bayesian networks' structures. Analyses of the results show that using experts' knowledge is more effective than other algorithms for increasing the accuracy of Bayesian network classification. A total of ten different diseases are used for testing, taken from the Machine Learning Repository datasets of the University of California at Irvine (UCI).

The proposed method improves both the computation time and accuracy of the classification methods used in this paper. Bayesian networks based on experts' knowledge achieve a maximum average accuracy of 87 percent, with a minimum standard deviation average of 0.04 over the sample datasets among all classification methods.

The proposed methodology can be applied to perform disease differential diagnosis analysis.

This study presents the usefulness of experts' knowledge in the diagnosis while proposing an adopted improvement method for classifications. Besides, the Bayesian network based on experts' knowledge is useful for different diseases neglected by previous papers.

We compare methods to measure comovement in business cycle data using multi-level dynamic factor models. To do so, we employ a Monte Carlo procedure to evaluate model performance for different specifications of factor models across three different estimation procedures. We consider three general factor model specifications used in applied work. The first is a single-factor model, the second a two-level factor model, and the third a three-level factor model. Our estimation procedures are the Bayesian approach of Otrok and Whiteman (1998), the Bayesian state-space approach of Kim and Nelson (1998) and a frequentist principal components approach. The latter serves as a benchmark to measure any potential gains from the more computationally intensive Bayesian procedures. We then apply the three methods to a novel new dataset on house prices in advanced and emerging markets from Cesa-Bianchi, Cespedes, and Rebucci (2015) and interpret the empirical results in light of the Monte Carlo results.