Flash presentation - Neural-based Analogical Reasoning for Link Prediction in Knowledge Graphs
Abstract: Knowledge graphs are used as data models and pivot datasets in many information systems and domains. However, their automatic or crowdsourced construction and maintenance raise completeness and correction issues, motivating the development of link prediction methods to complete them. In this seminar, I will introduce a method that performs link prediction via analogy detection.
10:30 - 11:00 Clara Thomas (Inria - PhD)
How to reduce the energy consumption of a industrial Pick and Place Robot
Abstract: Pick and Place robots represent approximately 48% of industrial robots worldwide. Yet, despite performing a relatively simple task—moving an object from one location to another—they consume disproportionately high amounts of energy. This raises critical concerns in terms of efficiency and sustainability in modern manufacturing. To address this inefficiency, we propose a three-fold strategy to significantly reduce their energy consumption. First, we explore architectural modifications to reduce inherent energy demands. Second, we optimize the robot’s geometry to minimize mechanical effort. Finally, we introduce a novel control approach based on open-loop actuation that eliminates the need for real-time computation or external control units. Together, these improvements provide a practical path toward simpler, more energy-efficient Pick and Place robots.
