Intelligent Information Processing

Mission

The Intelligent Information Processing (I2P) Lab was created in 2006 as a multidisciplinary research environment for developing research activities revolving around the design of smart computerized processing and analysis systems for a broad variety of information sources and applications such as environmental monitoring, biomedical and biometric systems, food quality control, multimedia management, document processing and software agents. To reach this ambitious objective, I2P gathers a team of young and active researchers with very different skills from the information and communication technologies (ICT) area. Moreover, it can rely on a rich and excellent network of national and international collaborations. Several contracts and more than one hundred scientific publications are the first very promising results achieved in these first years of I2P's life. There is, however, still a lot of work to be done.

Research Areas

I2P is developing research activities around different technological axes, which include:

• Image and signal processing methodologies for dealing with mono-, multi- and hyper-dimensional signals/images as well as multimodal signals/images. The main faced issues are adaptive filtering, segmentation and reconstruction of missing data. A particular emphasis is given to the design of transform-based techniques and statistical methods such as Markov random fields.
• Pattern recognition and machine learning deal with classification, regression and prediction problems in various application domains. Technologies based on support vector machines, artificial neural networks, kernel methods, statistical reasoning, data fusion as well as evolutionary computation and swarm intelligence have been designed with very promising results.
• Advanced negotiation and interaction mechanisms to support social environments through autonomous facilitators (software agents) capable of discovering and learning humans' behaviors. Data mining and collaborative filtering techniques are used by agents to discover no trivial relations among actions of different users and support their activities within the social environment (Implicit Culture framework).
• Data, content and knowledge representation technologies capable of overcoming  scalability issues characterize the existing approaches. The key idea is to view diversity as a valuable feature which must be maintained and exploited through a bottom-up strategy and not as a useless feature that must be absorbed in some general schema.
• Data Mining through classification techniques for huge datasets. The main problems addressed are those of increasing the accuracy with local models in usual classification tasks and in one-class classification tasks. In particular, the developed techniques exploit local models for classification achieving fast and scalable local versions of support vector machines.

Applications Areas

The continuous enhancement of the satellite remote sensor's characteristics has increased the importance of remote sensing in key real-world applications such as environmental monitoring (mapping, agriculture, forestry, inland and outland waters, and disaster prevention and monitoring). In particular, one of the reasons for the growing interest of numerous private and governmental end-users in the exploitation of remote-sensing data is represented by the higher quality and larger quantity of information that can be extracted through the analysis of remote-sensing images acquired over a given geographical area and characterized by improved spatial and spectral resolutions and reduced revisit time. I2P is particularly active in the development of technologies capable to better exploit such data and to follow the technological advances of remote sensors.

The automatic analysis of electrocardiogram (ECG) signals has received great attention from the biomedical engineering community since ECG provides cardiologists with useful information about the rhythm and functioning of the heart. I2P has been applying considerable efforts to design innovative systems for monitoring patients suffering from arrhythmia pathologies and prenatal cardiac activity.

During the last years, quantitative applications of infrared spectroscopy in various chemical fields including pharmaceutical, food and textile industries have grown drastically. Indeed, chemical analysis by spectrophotometry is very promising since it relies on a large number (hundreds and even thousands) of spectral data, which if suitably exploited can yield accurate estimations of the concentration of a chemical component of interest in a given product. I2P is developing advanced regression technologies for automating the quality control in food industries. The Implicit Culture framework has been applied to the development of knowledge management systems, the design and the run-time execution of service-oriented systems and in social environments where human users can interact through their mobile devices. Another challenge is the design of adaptive and, when necessary, self-adaptive knowledge systems as well as new tools for knowledge engineering and management, called managing diversity in knowledge by adaptation, and capable of harnessing, controlling and using the effects of emergent knowledge properties. The data mining techniques are applied to bioinformatics (miRNA target prediction), digital libraries (keyword extraction from documents) and data-driven spam filtering.

Members

Further Information

Projects