ANIMA

ANIMA - Adaptive Netting in Stream Data

The system ANIMA

In today's world, data is generated continuously (data streams) and in massive amounts. This data is often of different quality, structure, and granularity. From a contextual perspective, it is very difficult to find out whether there are some implicit hidden informational structures within this data that are appropriate for real-time applications.

The aim of ANIMA (Adaptive Netting In streaM dAta) [SHM04] is the efficient, real-time processing and analysis of streams of data. The project focuses on the generation of an artificial memory/brain that is able to represent existing associative relationships among informational entities in real-time.
Finding associative patterns in static databases has been an area of extensive research, and multiple approaches and solutions to the static problem have been presented in the past. (E.g. [AIS93], [AS94], [B98]) The major problem in these approaches is the combinatorial explosion of the search space and the research has therefore mainly focused on reducing this space. ANIMA is not subject to this problem because it uses an incremental approach for finding association rules. It is therefore particularly suited for time-critical/real-time applications.
Technically, ANIMA features a distributed architecture consisting of connectionist neural cells that follow the natural paradigm. That is, cells that are stimulated grow stronger, while cells that do not receive sufficient input are weakened and may die. Associations are represented by links interconnecting the neurons. These links are reinforced if a relationship between cells is detected in the data stream and degrade over time if the association is not observed for a while in the stream.
Currently, we focus on different research aspects like:

• The fusion of different/distributed data structures;
• The generation of communicative and interactive interfaces;
• The implementation of algorithms for the discovery of emergent patterns (e.g. hierarchical structures, temporal patterns, periodic patterns).
• Using ANIMA to analyse data gathered from the Wimbledon Tennis Championships. (see below)

Publication

2005

• C. Schommer, B. Schroeder: ANIMA: Associate Memories for Categorical data Streams. Proceedings of the 3rd International Conference on Computer Science and its Applications (ICCSA-2005). San Diego, USA.

2004

• C. Schommer: An incremental neural-based method to discover temporal skeletons in transactional data streams. To appear in Proceedings of the 5th Recent Advances in Soft Computing (RASC 2004), Nottingham, United Kingdom, 2004.
• C. Schommer: An SQL-like interface to retrieve associative patterns from neural skeletons. To appear in Proceedings 2004 International Conference on Advances in Intelligent Systems - Theory and Applications (AISTA 2004), Luxembourg, Luxembourg, 2004.
• Q. Sun, C. Schommer, A. Lang: Integration of Manual and Automatic Text Classification - A Text Categorization for Emails and Spams. Proceedings of the 27th German Conference for Artificial Intelligence, Ulm, Germany.
• C. Schommer: Incremental Discovery of Association Rules with Dynamic Neural Cells. Proceedings of the Workshop on Symbolic Networks. ECAI 2004, Valencia, Spain, 2004.

"ANIMA" is mentioned on: SEREBIF