KES-2012-Pattern-Recognition-Letters-issue-scope

De Grupo de Inteligencia Computacional (GIC)
Revisión del 14:06 7 ene 2013 de Manuel.grana (discusión | contribs.)
(difs.) ← Revisión anterior | Revisión actual (difs.) | Revisión siguiente → (difs.)
Title of the special issue
Innovative knowledge based techniques in pattern recognition
Submission
go to journal submission site [1]
select article type "special issue : KES"
Definition of issue’s scope

This special issue identifies the following knowledge related challenges in the current field of pattern recognition:

- Combining high (semantic) and low level information for pattern recognition.

Such needs appear in apparently unrelated fields like motion understanding, content based image retrieval (CBIR)

- Combining multi-modal information which can not be subject of conventional error measures, impeding the application of state-of-the-art error minimization approaches. The sources of information are heterogeneous, and its combination leads to unmeasurable data types. For instance, in the field of biomedical information processing, a wide diversity information sources may be used to reach a diagnosis that can be validated by the human operator.
- New views on uncertainty, ambiguity and data artifacts (i.e. noise, missing data) allowing to effectively deal with them in a systematic and integrative way.


These challenges are currently being addressed from a number of points of view. The development of semantic based pattern recognition systems, able to learn ontologies from data, even heterogeneous data, as well as of ontology based reasoning allow for the interplay between high-level semantics and low-level features. Hybrid systems combining ontologies with fuzzy systems allow introduce uncertain modeling and reasoning in the semantic domain. Lattice computing approaches allow seamlessly treatment of heterogeneous data through lattice theory, while allowing new more robust reasoning process diverging from the conventional statistics framework. Still, despite a long tradition of research and results, new classification algorithms are being extended to deal with heterogenous, ambiguous and artifact prone data. Also new ways of interactive development of systems, such as active learning or reinforcement feedback approaches can help to improve the efficiency of learning under scarce or expensive data collection. The special issue invites works on these areas showing the interplay between high level semantics and low level statistical and bio-inspired data processing techniques.

Specific technical topics
- Ontology reasoning and learning
- Fuzzy and probabilistic ontologies
- Lattice computing approaches
- Knowledge based adaptive approaches
- Active learning, reinforcement learning in qualitative spaces
- Advanced classification systems (ensembles, ELM, clustering)
- Interplay of high level semantics and low level machine learning
Applications
- Remote sensing
- Multimedia information processing
- Biometrics, such as face recognition
- Fusion of clinical and medical image information
- Content based image retrival
- Multimodal information processing


Submission deadline: 31st December 2012