Gurmukh Singh, visiting assistant professor in the Department of Computer and Information Sciences, recently published a research paper on pattern-recognition in the peer reviewed international journal, Physica A: Statistical Mechanics and its Applications.
The article, “Multifractal detrended fluctuation analysis of particle density fluctuations in high-energy nuclear collisions,” was featured in Vol. 424, pp. 25–33 of the April 15 issue.
The research is by Dr. Singh in an international collaboration with Dr. A. Mukhopahyaya, of North Bengal University (NBU) in Siliguri, India.
Singh investigated the formation of interesting patterns in big data sets collected in the 28Si and 32S induced heavy-ion interactions in experiments conducted from the two prominent national and international labs: the Brookhaven National Laboratory (BNL) in Upton and the European Center for Nuclear Research (CERN) in Geneva, Switzerland.
Singh and his collaborators developed the necessary algorithm to recognize hidden patterns through the analysis of big data sets collected from the two above cited experiments. The researchers employed a well-known technique of multifracts to recognize non-statistical significant patterns.
In scientific literature, a large number of multifractal parameters are computed, and then compared with a Monte-Carlo simulation based on the Ultra-relativistic Quantum Molecular Dynamics (UrQMD) model. The results of current big data analysis indicate that the single particle distributions in both experiments and in their respective UrQMD simulated results are multifractal in nature. The differences between the experiment and corresponding simulation, however, are not always very significant.
However, the results of their research are significantly different from those obtained by using other conventional methods of multifractal analysis. The present observations also indicate that the detrended multifractal analysis might be an efficient tool for characterizing the multi-particle emission patterns from such big data set. In future investigations, Singh noted the methodology will require some improvement so that it could differentiate between the non-statistical signal and the statistical noise.
Pattern-recognition has been historically investigated in a number of research fields such as computer science, psychology, psychiatry, ethology, cognitive science, traffic-flow etc. The technique of pattern-recognition has several advantages over its counterpart called pattern-matching. The pattern-recognition algorithms generally aim at to provide a reasonable answer for all possible inputs and to perform most-likely matching of the inputs, taking into account their statistical variation.
In contrast to pattern-recognition, pattern-matching is generally not considered a type of machine learning, although pattern-matching algorithms may sometimes are successful in obtaining a similar-quality output to the kind of results achieved by pattern-recognition algorithms.
For any additional questions or information, please contact Singh at singh@fredonia.edu.
