WAI › Schedule › 2018 › 09 apr

Title : Deep learning-based image recognition system for evolvable robots.

Presenter Jesus Benito-Picazo
Abstract Evolutionary robotics is one of the most popular filelds of study for many researchers and engineers. Inspired by the evolution processes followed by the living organisms along millions of years, it aims to produce individuals with better capabilities for adapting themselves to the environment changes in both structural and intellectual aspects. One of the mechanisms used by evolution for producing new individuals is to recombine the features of two individuals that currently exist through the copulation process. But for this to happen is very important that these individuals can locate the position of each other and to be as near as needed for the mating process to happen. With the consequent modifications, this process can be translated to robotics. Thus, our project aims to integrate a deep learning based image recognition system into the robot’s brain that allows them to locate and recognize themselves by processing the images provenient from an onboard camera. This way they will know what direction to move so they can be enough close to each other for the feature recombination process to be completed.

Title : Network Metrics for Assessing the Quality of Entity Links Between Multiple Datasets

Presenter Al Idrissou
Abstract Linking entities between datasets is a crucial step in data-integration in general, and in the use of multiple datasets on the semantic web in particular. A rich literature exists on different approaches to the entity linking problem, and a fair amount of tools is available for practical use. However, much less work has been done on how to assess the quality of such entity links once they have been generated by any of these tools. Evaluation methods for link quality are typically limited to either comparison with a ground truth (which is often not at one's disposal), manual work (which is cumbersome and prone to error), or crowd sourcing (which is not always feasible, especially if background information is required). Furthermore, the problem of link evaluation is greatly exacerbated for links between more than two datasets, because the number of possible links grows rapidly with the number of datasets. In this paper we propose a method to estimate the quality of such entity links between multiple datasets. We exploit the fact that the links between entities from multiple datasets form a network, and we show how simple metrics on this network of entity-links can reliably predict the quality of these links. We verify our results in a large experimental study using six datasets from the domain of science and innovation studies.