Andrei Doncescu

Non Negative Matrix Factorisation clustering capabilities; application on multivariate image segmentation

The clustering capabilities of the Non Negative Matrix Factorisation (NMF) algorithm is studied. The basis images are considered like the membership degree of the data to a particular class. A hard clustering algorithm is easily derived based on these images. This algorithm is applied on a multivariate image to perform image segmentation. The results are compared with those obtained by Fuzzy K-means algorithm and better clustering performances are found for NMF based clustering. We also show that NMF performs well when we deal with uncorrelated clusters but it cannot distinguish correlated clusters. This is an important drawback when we try to use NMF to perform data clustering.

Analytical and knowledge based approaches for a bioprocess supervision

The paper deals with supervision of biotechnological processes. Two main tasks are considered in this supervision scheme. The first one deals with Fault Detection and Isolation (FDI). The objective is to detect changes in the process dynamics using a model-based method. The second task concerns process physiological state recognition. A behavioral model is built using the expert knowledge along with clustering techniques applied to the measured data. The behavioral model under automata form is an heuristic model of the process under supervision.

Detection of the features of the objects in MR images using dynamic programming

Prostate cancer is the most frequently diagnosed cancer in males. It is also the second leading cause of cancer-related deaths in males after lung cancer. The treatment for patients with prostate cancer requires the diagnosis of the exact localization of the cancer. However, such a treatment remains a challenge, since the tumor biology of prostate cancer is still poorly understood. In the recent time, since developed imaging technologies, there are several methods have been developed to diagnose the prostate cancer based on the image. They allow understand more clearly the tumor biology of prostate cancer. In this paper, we propose to apply dynamic programming to detect the objects (tumors) in prostate images (MRI - magnetic resonance image) based on its features. Dynamic programming is one of the most powerful tools in optimization. We propose to use this technique to minimize the distance between active contours and the region representing the tumor. The challenge of this application is due to the noise and difficulties to capture the prostate tumor with high resolution and good contrast. The only solution to apply active contours which could interpolate between points converging on the region of interest. Dynamic programming takes into account the constraints set up by the user or by the environmental conditions related to image acquisition.

Default Logic for Diagnostic of Discrete Time System

Signaling pathways are seen as high criticalities in our understanding of mechanisms of biological functions. In this paper, we propose default logic for diagnostic of Discrete Time System (DTS) by focusing on automatic synthesis of the signaling pathways from factors within the cell. An essential component of this approach is using default logic to acquire facts about biological knowledge of intracellular communication. By choosing an adequate representation of biological knowledge, the reasoning is able to assign in acquisition of the facts and extract interactions necessary for the synthesis of the signaling pathways.

Image Medical Diagnostic Using Dynamic Programming

In recent years, the active contours model are used usually to determine objects in image. This model allows change coefficients in its formula to give interested results based on the energy optimization problems. There are many methods for solve them. In this paper, we try present the most fully using dynamic programming (DP) for the image medical diagnostic. This is among forceful and effective mathematical tools for solves the optimum problems. We present also technics set up automatically initial curve and constraints for decrease numbers of computation.

Ontologies applied in medicine: A digital library creation framework for health literacy

The work of this p aper is done within the context of European Project E! 8949 PrEmISES 1. One of the central pieces of PrEmISES[16] is the use of automatic tool generation algorithms for the creation of ontologies. The objective of this work is to extend the work from PrEmISES for raising the medical literacy of empowered patients by creating a digital library that contains relevant articles from different medical domain. It should be noted that a large number of persons are interested in finding relevant information about their condition. We have chosen to discuss in this research the automated generation of ontologies for the medical domain. By the use of ontologies that gives key terms from a domain, the creation of the digital library can be done by crawling from the Internet articles which are related to the profile of the user and that are based on the key terms of the ontologies. The articles from the digital library can be used by a patient to raise his / her knowledge and literacy related to his/her diseases.

Utilization of Default Logic for Analyzing a Metabolic System in Discrete Time

Metabolic pathways are seen as high criticalities in our understanding of mechanisms of biological functions. This article focuses on automatic synthesis of the metabolic pathways from factors in the domain of discrete time. An essential component of this approach is using default logic to acquire facts about biological knowledge of intracellular communication. By choosing an adequate representation of biological knowledge, the reasoning is able to assign in acquisition of the facts and extract interactions necessary for the synthesis of the metabolic pathways or signaling.

Reinforced Operators in Fuzzy Clustering Systems

Knowledge-based systems need to deal with aggregation and fusion of data with uncertainty. To use many sources of information in numerical forms for the purpose of decision or conclusion, systems are supposed to have tools able to represent the knowledge in a mathematical form. One of the solutions is to use fuzzy logic operators.

Spam filtering using Kolmogorov complexity analysis

One of the most irrelevant side effects of e-commerce technology is the development of spamming as an e-marketing technique. Spam e-mails (or unsolicited commercial e-mails) induce a burden for everybody having an electronic mailbox: detecting and filtering spam is then a challenging task and a lot of approaches have been developed to identify spam before it is posted in the end users mailbox. In this paper, we focus on a relatively new approach whose foundations rely on the works of A. Kolmogorov. The main idea is to give a formal meaning to the notion of information content and to provide a measure of this content. Using such a quantitative approach, it becomes possible to define a distance, which is a major tool for classification purposes. To validate our approach, we proceed in two steps: first, we use the classical compression distance over a mix of spam and legitimate e-mails to check out if they can be properly clustered without any supervision. It has been the case to highlight a kind of underlying structure for spam e-mails. In the second step, we have implemented a k-nearest neighbours algorithm providing 85% as accuracy rate. Coupled with other anti-spam techniques, compression-based methods could bring a great help in the spam filtering challenge.

Aircraft flight management with actuator major failure

This communication considers the case of total loss of hydraulic power by a transportation aircraft. The aim of this study is to propose control solutions to the design of new guidance function for the flight director system, or an emergency mode for the auto-throttle to control the thrust of a multi-engine transportation aircraft to make it return safely near ground level with acceptable landing conditions. The approach is based on a model of the flight dynamics involving all its major nonlinearities and couplings. Since after this major failure the aircraft is no more fully controllable, non linear inverse control technique is used to design control laws allowing the realization of a sequence of basic manoeuvres sufficient to turn achievable emergency landing. Some simulation results are displayed.