人工智能 | SCI期刊约稿信息7条

人工智能

Connection Science

Contribute to our special issue on Learning from Data Streams and Class Imbalance

全文截稿: 2018-03-31

影响因子: 0.867

CCF分类: C类

中科院JCR分区:

  • 大类 : 工程技术 - 4区

  • 小类 : 计算机：人工智能 - 4区

  • 小类 : 计算机：理论方法 - 4区

网址: http://www.tandfonline.com/toc/ccos20/current

With the wide application of machine learning algorithms to the real world, class imbalance and concept drift have become crucial learning issues. Applications in various domains such as risk management, anomaly detection, fraud detection, software engineering, social media mining, and recommender systems are affected by both class imbalance and concept drift. Class imbalance happens when the data categories are not equally represented, i.e., at least one category is minority compared to other categories. It can cause learning bias towards the majority class and poor generalization. Concept drift is a change in the underlying distribution of the problem, and is a significant issue specially when learning from data streams. It requires learners to be adaptive to dynamic changes.

Class imbalance and concept drift can significantly hinder predictive performance, and the problem becomes particularly challenging when they occur simultaneously. This challenge arises from the fact that one problem can affect the treatment of the other. For example, drift detection algorithms based on the traditional classification error may be sensitive to the imbalanced degree and become less effective; and class imbalance techniques need to be adaptive to changing imbalance rates, otherwise the class receiving the preferential treatment may not be the correct minority class at the current moment. Therefore, the mutual effect of class imbalance and concept drift should be considered during algorithm design.

The aim of this special issue is to bring together the original work from the areas of class imbalance learning and concept drift in order to solve the combined issue of class imbalance and concept drift. In order to advance the state-of-the-art on the combined issue, it is important to also advance the state-of-the art in each individual area. Therefore, this special issue encourages submissions not only on the combined issue, but also on these two areas themselves.

Topics of Interest  

The list of possible topics includes, but is not limited to:

(1) Research topics related to the combined issues of class imbalance and concept drift:
- Concept drift detection in imbalanced data streams.
- New data-level and algorithm-level approaches to dealing with class imbalance in non-stationary environments.
- Semi-supervised learning and active learning approaches to dealing with imbalanced data streams.
- Adaptive ensemble approaches for imbalanced data streams.
- Performance evaluation on imbalanced data streams in incremental and online learning scenarios.
- Case studies and real-world applications dealing with both class imbalance and concept drift.

(2) Research topics related to class imbalanced learning:
- Data-level and algorithm-level techniques for imbalanced data.
- Ensemble learning approaches for imbalanced data.
- Cost-sensitive and cost-free learning approaches.
- Imbalanced data with multiple classes or multiple labels.
- Semi-supervised class imbalance learning.
- Case studies and real-world applications dealing with class imbalanced data.

(3) Research topics related to learning in the presence of concept drift:
- Passive and active approaches to dealing with concept drift.
- Concept drift detection methods.
- Chunk-based and online learning approaches for non-stationary environments.
- Approaches to dealing with recurring concepts.
- Adaptive ensemble approaches.
- Semi-supervised learning in non-stationary environments.
- Case studies and real-world applications involving concept drift.

人工智能

Neurocomputing

Special Issue on Protocol-Based Performance Analysis of Artificial Neural Networks and Their Applications

全文截稿: 2018-05-15

影响因子: 3.317

CCF分类: C类

中科院JCR分区:

  • 大类 : 工程技术 - 2区

  • 小类 : 计算机：人工智能 - 3区

网址: http://www.journals.elsevier.com/neurocomputing/

Recent years have seen a growing number of publications reporting on neural networks (NNs) due to their extensive applications in a broad range of areas such as, repetitive learning, classification of patterns, nonlinear control, adaptive control, image processing, and so forth. For real-world engineering, complex dynamics coming from multiplicative noises, data missing and communication delays are commonly unavoidable in various applications of NNs. These complex dynamics have a major impact on the dynamical behaviour and the precision of state estimation, and can be further regarded as a crucial source of negative effects such as periodic oscillation, divergence and even chaos. As such, to date, much research effort has been devoted to the dynamic performance analysis, and a variety of efficient approaches have been proposed in the published literature.

It is worth pointing out that the complex dynamics are much more than those already studied. Specially, frequent data communication is necessary for the application of NNs in networked engineering systems. Due to the limited channel bandwidth, various communication protocols are employed to orchestrate the transmission order of networked nodes to avoid data collisions. For instance, under the event-based protocols, the information processing is triggered when the changing degree of information or its derivatives break through the predetermined level. Unfortunately, applying communication protocols could lead to asynchronous coupling, periodic switches, accumulated delays, or unknown-but-bounded disturbances, and so forth. In this case, it would be interesting to examine 1) how communication protocols have substantial impacts on the dynamical behaviour, 2) how can engineers design the schemes of state estimation and synchronization control of coupled NNs, and 3) how can engineers implement the nonlinear control, adaptive control, or other engineering tasks based on NNs under various communication protocols. As such, protocol-based performance analysis of NNs has been attracting an ever increasing research interest.

Research on NNs under communication protocols has already become vitally important for control engineers, mathematicians, and computer scientists to analyze and interpret the complex dynamical behaviour of NNs, and effectively implement the NN-based engineering tasks. This special issue aims to bring the latest approaches to investigate NN its analogue under communication protocols in a quantitative way.

The list of possible topics includes, but is not limited to:

- Protocol-based state estimation of NNs

- Protocol-based synchronization analysis of coupled NNs

- Finite-time performance analysis and synthesis for NNs under communication protocols

- NN-based control under various communication protocols

- Protocol-based state estimation of complex networks

- Protocol-based synchronization analysis of complex networks

- Protocol-based distributed filtering of sensor networks

人工智能

Neurocomputing

Special Issue on Deep Learning for Medical Image Analysis

全文截稿: 2018-05-31

影响因子: 3.317

CCF分类: C类

中科院JCR分区:

  • 大类 : 工程技术 - 2区

  • 小类 : 计算机：人工智能 - 3区

网址: http://www.journals.elsevier.com/neurocomputing/

In medical image analysis, the accurate diagnosis of a disease depends on two aspects: medical image acquisition and medical image interpretation. Medical image acquisition has grown substantially over recent years, with devices acquiring data at faster rates and increased resolution. The medical image interpretation has only recently begun to benefit from computer technology, and most interpretations on medical images are performed by physicians. However, image interpretation by humans is limited due to its subjectivity, large variations across interpreters, and fatigue. Many diagnostic tasks require an initial search process to detect abnormalities, and to quantify measurements and changes over time. Computerized tools, specifically image analysis and machine learning, are the key enablers to improve diagnosis, by facilitating identification of the findings that require treatment and to support the expert’s workflow. Among these tools, recent advances in machine learning, especially with regard to deep learning, are helping to identify, classify, and quantify patterns in medical images. At the core of these advances is the ability to exploit hierarchical feature representations learned solely from data, instead of features designed by hand according to domain-specific knowledge. Deep learning is rapidly becoming the state of the art, leading to enhanced performance in various medical applications.

The special issue seeks for original contribution of works which addresses the challenges from the deep learning techniques for medical image analysis. Papers on pure medical imaging would be out of the scope of this special issue. The list of possible topics includes, but not limited to:

- Convolutional Neural Networks (CNNs) in diseases diagnosis

- CNNs in medical image segmentation, fusion, shape modeling, and etc.

- Other deep learning networks in diseases diagnosis

- Unsupervised deep learning methods in medical image analysis

- Supervised deep learning methods in medical image analysis

- Transfer learning and fine-tuning methods in medical image analysis

人工智能

Computer Vision and Image Understanding

Special Issue on “Adversarial Learning in Computer Vision”

全文截稿: 2018-06-15

影响因子: 2.498

CCF分类: B类

中科院JCR分区:

  • 大类 : 工程技术 - 3区

  • 小类 : 计算机：人工智能 - 3区

  • 小类 : 工程：电子与电气 - 3区

网址: http://www.journals.elsevier.com/computer-vision-and-image-understanding

Generative Adversarial Networks (GANs) have been a breakthrough in machine learning and since their introduction in 2014, they have quickly become a fundamental asset in modern computer vision and deep learning. New adversarial models are proposed at an accelerating pace that increase the level of realism of synthesized data and/or uncover missing explanations for its workings and failures. Beside generation capabilities, adversarial learning techniques provide a powerful framework for using unlabeled data to train machine learning models, rising as one of the most promising paradigms for unsupervised learning.

However, due to the novelty of these approaches, we need to develop principles to understand them better, from both theoretical and empirical perspectives, as well as expand their applications to tackle problems with real-world complexity (e.g., image and video content understanding, motion analysis, super-resolution, image translation, etc.).

Given the above premises, the objective of this special issue is: a) to provide a comprehensive overview of the most recent GAN models and architectures; b) to provide means for explaining theoretically and empirically GANs; and c) to present and report new applications of adversarial models for computer vision.

Submissions are encouraged, but not limited, to the following topics:

- Comparative analysis of GAN models

- Theoretical models and/or theory-grounded metrics for performance assessment

- Explainable generative adversarial models

- Adversarial learning to improve traditional training approaches

- Adversarial-based unsupervised learning frameworks

- Generation of synthetic data supporting the creation of methods in domains with limited data (e.g., medical image analysis)

- Application of GANs to traditional computer vision problems:

- 2D image content understanding: classification, detection, semantic segmentation

- Video dynamics learning: motion segmentation, action recognition, object tracking

- 3D scene understanding: stereo vision, reconstruction, 3D object detection, depth assessment

- New applications of adversarial learning in vision

人工智能

International Journal of Approximate Reasoning

A Special Issue of International Journal of Approximate Reasoning on Decision Making under uncertainty and imprecision

全文截稿: 2018-11-30

影响因子: 2.845

CCF分类: B类

中科院JCR分区:

  • 大类 : 工程技术 - 2区

  • 小类 : 计算机：人工智能 - 2区

网址: http://www.journals.elsevier.com/international-journal-of-approximate-reasoning/

The choice between several alternatives in a decision-making problem can be rendered difficult by the existence of uncertainty in the consequences of these alternatives. The standard approach to this issue is to model this alternative by means of probability theory, and to consider then a stochastic order, such as expected utility or stochastic dominance.

However, when the probabilistic information available is vague or scarce, the elicitation of a precise probability model can be difficult, and its use, questionable. In that case, it is possible to make use of tools from Imprecise Probability Theory, such as fuzzy measures, belief functions, possibility measures or lower/upper previsions, to model our uncertainty. Moreover, in some cases the incomplete information about the consequences of the alternatives can be tackled by means of the Theory of Fuzzy Sets.

This Special Issue aims at gathering significant advances in decision making problems with partial information. We encourage both theoretical and practically oriented papers. High-quality papers introducing novel approaches, improved methods or outstanding applications are welcome.

Topics of interest include, (but are not limited to):

- Decision making with imprecise probabilities.

- Connections between game theory, fuzzy measures and imprecise probabilities.

- Stochastic orderings with imprecise information.

- Multivariate modelling with imprecise probabilities.

- Choice functions.

…

人工智能

Applied Soft Computing

Virtual Special Issue: New Paradigms, Trends and Applications of Machine Learning and Soft Computing in Cyber-Physical Systems

全文截稿: 2018-12-31

影响因子: 3.541

中科院JCR分区:

  • 大类 : 工程技术 - 2区

  • 小类 : 计算机：人工智能 - 2区

  • 小类 : 计算机：跨学科应用 - 2区

网址: http://www.journals.elsevier.com/applied-soft-computing/

The last few years have witnessed an explosion of research activity around the so-called Cyber-Physical Systems (CPS), conceived as architectures, protocols, standards, platforms, services and applications with a high level of integration and interaction of software and physical components. This broad definition embraces a myriad of technologies from different disciplines (such as aimed at bridging the gap between the physical and the digital worlds, moving at different temporal and spatial scales and with very diverse, yet complementary, capabilities in regard to ubiquity, interactivity, cognition, self-configurability, dynamicity, usability and adaptability. Several sectors have lately exploited the enormous benefits foreseen for CPS, from Energy (smart grids, energy efficiency for buildings) to Industry 4.0 (smart robotics, mechatronics) through Health (body nets, robot surgery), Operations Research (firefighting, disaster missions) or Transport (collision avoidance, driving efficiency), among many others. It is not in vain that the market potential predicted for CPS is huge, with billions of dollars’ worth revenues expected for sectors implementing these systems. As to mention, forecasts for the intelligent sensor market, a core part of the CPS technology spectrum, are foreseen to be a $10.5 billion industry in 2020 (source: Deloitte).

From a technical perspective, CPS can be regarded as a rich substrate, where many different technologies collide with no clear means for their implementation, integration, and coordination. The integration itself poses a very challenging scenario in what refers to the communication between components and functionalities coming from radically different disciplines (e.g., computer science and mechanical engineering), with differing programming languages, constraints and requirements. Technologies such as Virtual Reality, Internet of Things, Ultra-Reliable Low-Latency communications, Mechatronics, Embedded Systems or Data Analytics lie at the core of CPS, and are called to revolutionize all the aforementioned sectors, by enabling smart CPS capable of capturing information from their interaction with the physical environment and mining it towards an increased level of intelligence (e.g., better decision making or increased self-* capabilities).

However, several technological challenges stem from the particularly complex characteristics of CPS, far beyond the aforementioned integration of portfolios of diverse technologies. To begin with, overly involved are those barriers related to the inference of knowledge from the captured data, which can be implemented partly or fully at the sensors or, more usually, remotely queried upstream from the Cloud subject to the latency requirements of the application at hand. All in all, CPS span a plethora of paradigms in regards to its deployment, communication, organization, resource allocation, management, data collection, fusion, aggregation, analysis and human interaction, closely related to research trends such as Fog Computing, Tactile Internet, Data Science, Optimization and other critical enablers. Computational Intelligence (in particular, Machine Learning and Soft Computing techniques) are promising enablers of the intelligence and self-learning capability required for the CPS to succeed in tackling the above challenges.

Authors are invited to submit their original works focusing on how CPS can benefit from its synergy with Soft Computing techniques and Machine Learning models, with an emphasis on evidences of the practicability of the reported findings. Topics of interest for this special issue include, but are not limited to, the following:

- Novel Soft Computing techniques and their application to problems related to CPS, such as distributed predictive modelling, hybrid optimization techniques, online learning over data streams, concept drift adaptation, automated model construction, large-scale deployment of Soft Computing techniques, collaborative reasoning and weakly/semi-supervised learning, among others

- Data analytics and scalable/parallel/distributed computing algorithms for CPS

- Artificial Intelligence (AI) as a service (AIaaS) for CPS

- Energy efficiency paradigms for CPS tackled via Soft Computing and Machine Learning

- Distributed computing, data fusion and aggregation over large-scale CPS

- Predictive and clustering models for CPS self-configuration, self-resilience and self-autonomy

- Optimization algorithms for optimal sensor actuation

- Autonomic computing, inference of human patterns, analysis, monitoring, and situation alertness in CPS

- Federated learning, collaborative machine learning and distributed AI for large scale CPS

- Soft Computing techniques to enable ultra-reliable, low-latency applications in CPS scenarios.

Only submissions where novel Soft Computing techniques are applied to CPS scenarios as the ones described above will be considered for further editorial processing.

人工智能

Pattern Recognition Letters

Virtual Special Issue on “Recent Advances in Statistical, Structural and Syntactic Pattern Recognition”

全文截稿: 2019-01-31

影响因子: 1.995

CCF分类: C类

中科院JCR分区:

  • 大类 : 工程技术 - 3区

  • 小类 : 计算机：人工智能 - 4区

网址: http://www.journals.elsevier.com/pattern-recognition-letters/

Statistical, structural and syntactic pattern recognition are the most classical and important branches of pattern recognition research. They provide the fundamental theory and methods for a number of research areas in pattern detection, machine learning, computer vision, and data mining, leading to successful applications in many tasks such as handwriting and face recognition, image classification, video processing, and so on.

The study of statistical pattern recognition covers all stages of an investigation from problem formulation and data collection through to discrimination and classification, assessment of results and interpretation. It normally assumes that object samples are represented as feature vectors of fixed dimensionality, to which statistical techniques are applied for analyzing pattern variations. On the other hand, the study of syntactic & structural pattern recognition represents an object by a variable-cardinality set of symbolic, nominal features. This allows for representing pattern structures, taking into account more complex interrelationships between attributes than is possible in the case of numerical feature vectors of fixed dimensionality. As a result, the studies of statistical pattern recognition and syntactic & structural pattern recognition complement each other and make a comprehensive characterization of various pattern recognition problems.

Facing the fast development and wide applications in these areas, since 1996, the IAPR TC-1 (Statistical Pattern Recognition Techniques) and TC-2 (Structural and Syntactical Pattern Recognition) have sponsored a series of IAPR joint International Workshops on Statistical Techniques in Pattern Recognition and Structural and Syntactic Pattern Recognition (S+SSPR). S+SSPR is an international forum for exchanging ideas and discovering fancies in the fields of pattern recognition.

The proposed special issue will accept contributions significantly extended from papers presented in S+SSPR 2018, which is going to be held in Beijing during Aug. 17-19, 2018. It will also be a thematic issue open to the whole pattern recognition community. The scope of this special issue will be the same as S+SSPR conference, which ranges from structural based pattern recognition, statistical pattern recognition, and graphical model, to applications of deep learning, computer vision and data mining. The topics include, but are not limited to:

• Structural Matching
• Statistical Classification and Prediction
• Syntactic Pattern Recognition
• Graph-theoretic Methods
• Ensemble Methods
• Graphical Models
• Metric Learning
• Subspace Learning
• Structural Text Analysis
• Stochastic Structural Models
• Applications in statistical, structural and syntactic based pattern recognition methods

下载Call4Papers App，获取更多详细内容！