Cloud Roboticsの研究者であるProf. Peter SincakとDr. Gergely Magyarをお迎えして
場所：B4棟 4階 E420室
講演者：Prof. Peter Sincak and Dr. Gergely Magyar
所属：Technical University of Kosice, Slovakia
Center for Intelligent Technologies
講演題目：Towards Cloud-based Computational Intelligence
The talk will focus on the coherent research trends in Computational Intelligence and Cloud Computing trying to answer the question:
Does cloud computing affect computational intelligence and its research thinking?
This question is connected to the concept of Remote Brain introduced by Prof. Inaba (University of Tokyo) in 1997 which is supported by recent advancements in telecommunication technologies adapting the above-mentioned idea to the Cloud Robotics phenomena. This approach is fully compatible with the concept of Industry 4.0 and Reality 2.0. The talk will present the methods how to design, implement and use such systems in the cloud environment by multiple users. In particular, it will focus on AI bricks towards ‘Intelligence as a Service (IaaS)’. In general, these software components utilize conventional computational intelligence approaches and offer those to multiple users. Good examples are TensorFlow by Google, Watson by IBM or Azure Machine Learning by Microsoft. Our interest is to create a domain-oriented IaaS for Ambient Assisted Living.
The next part of the talk will deal with a concrete AI brick for classification using fuzzy sets. Firstly, the domain of pattern recognition and classification will be presented. Then the Membership Function ARTMAP classification approach, whose goal is to obtain a lower classification error, will be explained in detail. Its performance will be compared with another method for classification, namely the Cumulative Fuzzy Class Membership Criterion, using various data sets. This novel classification concept is a result of our collaboration with Prof. Hartono (Chukyo University).
The last part of the talk will be about the connection of reinforcement learning and cloud computing. The presentation will cover the advantages of such a fusion and its use in social human-robot interaction. In addition to this a concrete example of social HRI will be presented on video, where the robot acted as a coach for cognitive stimulation therapy in an elderly care facility. In this case, reinforcement learning and cloud computing was used to learn social action selection from the operator of the robot and by doing so increasing the robot’s level of autonomy and creating a personalized behavior for each patient.