The title and abstract of talk are attached at the bottom of this message.
The lecture is primarily targeted to the SCCP class (O3-019 A Peek Inside Computers), but we will make it open to general public.
Here are some information regarding the University of Porto and its relationship to Aizu:
- Aizu and Porto have an exchange program agreement.
- Dr Pedroso from Porto visited Aizu and gave us special lectures Link
- I myself visited Porto and introduced Aizu to Porto Link
- One of our students, Kazuaki Takahashi, visited Porto in 2009 and Dr Pedroso took care of him. Link Link
- A student from Porto, Rui Jorge Rei, was staying with us from January to April 2010 under JASSO scholarship support. Dr Pedroso is his supervisor. Link
Date and TimeDec 5, 2014, from 1810 to 1940
VenueResearch Quadangle 275-S4
TitleIntroduction to Inductive Logic Programming
AbstractInductive Logic Programming (ILP) is a powerful but often neglected machine learning technique for extracting knowledge from observations; it consists of building a generalized logic model which explains a set of facts, to a degree of accuracy. The main advantage of ILP is its capability to provide classifiers that are interpretable by humans, which makes it popular in different scientific domains. This research area is quite relevant in several field such as knowledge acquisition, inductive program synthesis, inductive data engineering, and knowledge discovery in databases. Much work has been done on finding good models in ILP, but there is still a vast scope for improvements, in particular with respect to attaining faster execution times; because ILP systems must deal with enormous search spaces, they usually incur in non-trivial query checking times. To tackle this issue, it is possible to take advantage of the concurrent sections of the ILP algorithm and paralellize them so as to reduce execution time significantly; a strong case for the parallelization of the search space can be made: on the one hand, there are several sections of the ILP algorithm which could be executed concurrently without adding much to the complexity of the code; on the other hand, the fact that datasets are becoming too large to be stored in one s memory invites to the use of several machines connected in heterogeneous configurations such as clusters, grids or clouds.machine
Speaker's BioJoana Corte-Real is a PhD student of the joint Doctoral Program of Universities of Minho, Aveiro and Porto in Computer Science (MAP-i). She obtained her MSc degree from the Faculty of Engineering of University of Porto in 2013, in Computers and Electric Engineering. Her current interests are parallel and distributed computing and machine learning, in particular Inductive Logic Programming techniques. She developed a parallel version of a MapReduce construct for Prolog that obtains linear speed-ups in multicore machines. Currently, she has been researching on how to extend this implementation for hybrid shared and distributed memory architectures and one of the applications of this work is on Inductive Logic Programming. On her free time, she enjoys horse riding and takes german classes; she is also keen on traveling and meeting new people.
Post a Comment