DCCI 2013 Abstracts


Short Papers
Paper Nr: 4
Title:

Stability-aware Cognitive Packet Network Routing Protocol for MANET

Authors:

A. Alharbi, A. Aldhalaan and M. Alrodhaan

Abstract: The research introduces a new Stability-Aware Cognitive Packet Routing Algorithm to provide Quality Of Service (QoS) to packets in Mobile Ad hoc networks (MANET) through long-lived, short delay routes. It extends the work on Ad-hoc Cognitive Packet Network (AHCPN) to adapt it to the MANET environment. It uses Smart Packets (SP) to find and maintain paths that provide specific QoS goals to each packet. Smart Packets move through the network collecting information to be able to make decisions by “learning” from the experience of previous SP’s of the same QoS goals. The routing algorithm constructs Random Neural Networks (RNN) at each node with a neuron for each output link in order to make a routing decision. The algorithm uses Computational Intelligence to make routing decisions with high probability to stable nodes in th e network. A feedback system is subsequently used based on Reinforcement Learning through Acknowledgments to update network information in RNN’s. The proposed algorithm defines a combined routing goal function that considers both the associativity of the node for stability and path delay to result in short, long-lived paths. The degree of associativity of a node is determined via low frequency periodic beacons to identify itself to its neighbors.
Download

Paper Nr: 5
Title:

Automated Classification of Haematopoietic Compartments in the Human Bone Marrow using Reservoir Computing

Authors:

Philipp Kainz, Harald Burgsteiner, Helmut Ahammer and Martin Asslaber

Abstract: Background. Histomorphometry of haematopoiesis in the human bone marrow is a mandatory element in lots of daily diagnosis processes in pathology. The determination of relative quantities of the haematopoietic compartments is currently performed visually by the individual pathologist using conventional microscopy. Hence, intra- and inter-observer variability is unavoidable, but standardized quantitative methods are not available yet. Standard image processing methods are limited when it comes to automated classification of objects within a histological image but methods and paradigms of Computational Intelligence (CI) have the potential to overcome these barriers. Specific Aims. The proposed PhD project is intended to develop and implement a machine learning system for the automated quantification of objects in histological images. The major tasks are the development of a classifier based on the reservoir computing paradigm for automated pattern recognition and classification as well as its prototypical software implementation. Research Methods. Histological sections of human bone marrow are stained using histological standard techniques. Experienced pathologists will label the haematopoietic compartments in a software system and the data sets for the classifier are generated. We are going to train the algorithm on the labeled image patches in order to distinguish different cell classes. Expected Results. This classification system will contribute to the progress in digital pathology in terms of decreasing the overall intra- and inter-observer variability in the diagnostics of human bone marrow specimen. Furthermore, we emphasize the potential of CI algorithms in medical image analysis and pattern recognition.
Download

Paper Nr: 6
Title:

Advanced Learning Techniques for Chemometric Modelling

Authors:

Carlos Cernuda, Edwin Lughofer and Erich Peter Klement

Abstract: The European chemical industry is the world leader in its field. 8 out of the 15 largest chemical companies are EU based. Furthermore, 29 % of the worldwide chemical sales originate from the EU. These industries face future challenges such as rising costs and scarcity of raw materials, an increase in the price of energy, and an intensified competition from Asian countries. Process Analytical Chemistry represents one of the most significant developments in chemical and process engineering over the past decade. Chemical information is of increasing importance in today's chemical industry. It is required for efficient process development, scale-up and production. It is used to assure product quality and compliance with regulations that govern chemical production processes. If reliable analytical information on the chemical process under investigation is available, adjustments and actions can be undertaken immediately in order to assure maximum yield and product quality while minimizing energy consumption and waste production. As a consequence, chemical information has a direct impact on the productivity and thus competitiveness, and on the environmental issues of the respective industries. Chemometrics is the application of mathematical or statistical methods to chemical data. The International Chemometrics Society (ICS) offers the following definition: “Chemometrics is the science of relating measurements made on a chemical system or process to the state of the system via application of mathematical or statistical methods”. Chemometric research spans a wide area of different methods which can be applied in chemistry. There are techniques for collecting good data (optimization of experimental parameters, design of experiments, calibration, signal processing) and for getting information from these data (statistics, pattern recognition, modeling, structure-property-relationship estimations). In this extense list of tasks, we are focused on calibration. Calibration consists on stablishing relationships, i.e. chemometric models, between some instrumental response and chemical concentrations. The usual instrumental responses come from the use of spectrometers, because they allow us to get a lot of on-line cheap data in a non-destructive way. There are two types of calibration, univariate or multivariate calibration, depending on the use of only a single predictor variable or several ones. The current instalations provide us with thousands of variables and thousands of samples, thus more and more new sophisticated techniques, which are capable to handle and take advantage of this tsunami of data, are required. Our goal is to provide the analytical chemistry community with modern and sophisticated tools in order to overcome the incoming future challenges.
Download

Paper Nr: 9
Title:

Multiobjective Memetic Algorithms applied to University Timetabling Problems

Authors:

Nuno Leite, Fernando Melício and Agostinho Rosa

Abstract: The present Ph.D. Thesis Proposal focus the study and implementation of efficient Multiobjective Memetic Algorithms and its application to University Timetabling Problems. These problems will also be studied and solved in a many-objective framework which impose new problems such as quality of approximated Pareto Front, solution diversity, pertinency of solutions to the decision maker, visualization of chosen solutions, and among others.
Download

Paper Nr: 10
Title:

The Consumer Prototype - Explaining the Underlying Psychological Factors of Consumer Behaviour with Artificial Neural Networks

Authors:

Max N. Greene

Abstract: Consumer behaviour is examined using artificial neural networks as a method of analysis to identify the underlying psychological factors that influence consumer choice. Artificial consumer prototype is developed and consequently studied using supervised and unsupervised neural networks. A number of network architectures are constructed and optimized for comparative purposes. Learning obtained using artificial agents is interpreted and subsequently propagated towards human consumer behaviour. Philosophical issues including the network structure interpretation and appropriateness of using artificial agents to explain human behaviour are discussed.
Download