INVITED TALKS
Talks by Prof. Zhi-Hua Zhou, Prof. Keith C. C. Chan, and Dr. Jordi Herrera-Joancomartí will be given in MDAI 2011. Information follows.
ABSTRACTS OF INVITED TALKS
Prof. Zhi-Hua Zhou
(National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China)
Cost-Sensitive Learning
Abstract:
In conventional classification settings, the classifiers generally try to maximize the accuracy or minimize the error rate, both are equivalent to minimizing the number of mistakes in classifying new instances. Such a setting is valid when the costs of different types of mistakes are equal. In real-world applications, however, the costs of different types of mistakes are often unequal. For example, in intrusion detection, the cost of mistakenly classifying an intrusion as a normal access is usually far larger than that of mistakenly classifying a normal access as an intrusion, because the former type of mistakes will result in much more serious losses. In this talk we will introduce how to handle unequal costs in classification.
Prof. Keith C. C. Chan
(Department of Computing, Hong Kong Polytechnic University, Hong Kong, China)
Evolving Graph Structures for Drug Discovery
Abstract:
Computer-Aided Drug Discovery (CADD) is concerned with the use of
computational techniques to determine drug structures with certain desirable properties.
Evolutionary algorithms (EAs) have been proposed to evolve drug molecules by mimicking
chemical reactions that cause the exchange of chemical bonds and components between
molecules. For these EAs to perform their tasks, known molecular components, which can
serve as building blocks for the drugs to be designed, and known chemical rules, which
govern chemical combination between different components, have to be introduced before
an evolutionary process can take place. To automate drug molecular design without such
prior knowledge and constraints, we need a special EA that can evolve molecular graphs
with minimal background knowledge. In this talk, we present one such EA that can evolves
graph structures used to represent drug molecules. We show how molecular fingerprints can
be used to evaluate the .fitness. of an evolved drug structure obtained at each generation
during the evolutionary process. We also show how the discovering of privileged structures
in many drug molecules and the use of ligand docking and binding affinity can be used as
alternatives for fitness evaluating in an EA for drug design. We show how the results
obtained using the proposed EA may lead to a promising approach for CADD.
Dr. Jordi Herrera-Joancomartí
(Department of Information and Communications Engineering, Universitat Autònoma de Barcelona, Catalonia, Spain)
Online social honeynets: trapping web crawlers in OSN
Abstract: Online social networks (OSNs) have become an important part of people's everyday communication. With millions of individuals who use OSNs to share all kinds of contents, privacy concerns of how all this content is managed have arisen. Content shared in an OSN varies from trivial text messages to compromising photographs but, in either of those cases, users expect to control their shared data with their profile's visibility configuration. In addition to this personal data, users in OSNs create relationships that can also be considered sensitive data from themselves. Moreover, the discovering of these relationships can also produce other data revelation, what makes link privacy an important issue to preserve in social networks.
OSN information can be obtained by crawling the profiles of users in the network. Web crawlers are complex applications that explore the Web with different purposes and they can be configured to crawl OSN to obtain both user and link information. In this talk, we will review the basic architecture of a web crawler and then we will present the concept of online social honeynet. Online social honeynets are, much like traditional honeynets, a set of users in the social network whose objective is defend the network from attackers that want to retrieve information from the network. Also, like traditional honeynets, online social honeynets consist of a set of users that appear to be part of the network with information of value to the attackers but they are actually isolated and monitored. We will show the properties that such social honeynets should have in order to trap web crawlers and we will discuss the opportunities and drawbacks that such technique offers.
IIIA - Institut d'Investigació en Intel·ligència Artificial 