CfP: special issue on Information Fusion in Learning from Data Streams

28 views
Skip to first unread message

Michal Wozniak

unread,
Feb 5, 2016, 4:52:13 AM2/5/16
to concept drift
The Information Fusion Journal (http://www.journals.elsevier.com/information-fusion/) is planning a special issue on Information Fusion in Learning from Data Streams.


Deadline for Submission: July 15, 2016


Contemporary computing systems cannot be restricted to processing canonical, stationary data. We live in era of massive amount of information being generated in real-time. Many real-life domains, like internet traffic, bank operations, mobile connections, or social media became a highly important source of data for mining valuable knowledge. In such applications data arrives continuously in the form of data stream. To be able to extract any meaningful information from it, we require algorithms that are able to handle real-time analytics, while regularly updating their model when new data is available. One must have in mind, that such a scenario imposes additional constraints such as processing time, memory usage or recovery rate. Another important issue is the problem of non-stationary environment, as data stream may change its nature over time with the occurrence of shifts and drifts in its characteristics. There is also a lot to be done on how to evaluate algorithms for mining data streams. Another difficulty lies in the  fact that such  data often originates from more than a single source. Data can be generated from many different sources and provided at the same time, offering us an access to varied information to feed to our classifiers. Fusion of data streams have been so far mainly researched from the database perspective. Yet this vital issue requires a proper addressing from the learning systems point of view. Contemporary applications like social media provide us with massive but highly diverse and unstructured information in form of ubiquitous data streams. Therefore, proposing fusion-based learning systems for such scenarios is crucial. Additionally, in reality we may face a more pessimistic scenario – data can be noisy, of differing certainty and quality or even contradictory to each other.

This special issue aims at gathering recent developments in information fusion for learning from streaming and evolving data. We welcome contributions presenting novel and impactful research on this topic. We aim at gathering research articles that will address some of the important topics in learning from high-speed and changing streams, point out new directions, discover new strong and weak points of methods or present efficient and complex applications to vital real-life areas.

Manuscripts (which should be original and not previously published either in full or in part or presented even in a more or less similar form at any other forum) covering unpublished research that clearly delineate the role of information fusion in the context of learning from data streams are invited.

The manuscript will be judged solely on the basis of new contributions excluding the contributions made in earlier publications. Contributions should be described in sufficient detail to be reproducible on the basis of the material presented in the paper and the references cited therein.

Topics appropriate for this special issue include (but are not necessarily limited to):

  1. Mining evolving and heterogeneous data streams.
  2. Online, evolving, autonomous and self-adapting learning systems.
  3. Ensemble methods for streaming and multi-source problems.
  4. Change detection and reacting to concept drift.
  5. Mining complex streams, such as imbalanced, uncertain, or multi-label data.
  6. Learning from high-speed and massive data streams.
  7. Managing dynamic ensemble structures in the presence of drifting data, pruning and combination methods.
  8. Analyzing heterogeneous data streams with varying or mixed data types.
  9. Deep learning from streaming and evolving datasets.
  10. One-class classification in non-stationary environments.
  11. Collaborative forgetting of outdated data.
  12. New methods for evaluating algorithms for data streams.
  13. Handling non-stationary nature of recognition problems: changes in class structures, new data sources etc.
  14. Emerging real-life applications such as activity recognition, network intelligence or business analytics.
  15. Mining streaming data originated from wireless sensor networks and Internet of Things (IoT)
  16. Mining social media streaming data, and network dynamics
  17. New hardware solutions for data stream mining
  18. Scalable software solutions for data stream mining
  19. Manuscripts should be submitted electronically online at http://ees.elsevier.com/inffus. 
The corresponding author will have to create a user profile if one has not been established before at Elsevier. 

To ensure that all manuscripts are correctly identified for inclusion into the special issue, it is important that authors select “SI:INFFUS from data Streams".

Guest Editor(s)

Bartosz Krawczyk
Wrocław University of Technology, Poland

Michał Woźniak
Wrocław University of Technology, Poland

Mohamed Medhat Gaber
Robert Gordon University, UK

Krzysztof Cios
Virginia Commonwealth University, U.S.A.


Reply all
Reply to author
Forward
0 new messages