Intel Lab, Neuromopic Computing Improvement of Sustainable Interactive Learning for Robots

Intel Lab announced on the 2nd that it has introduced a new approach to neural network-based object learning in cooperation with the Italian Institute of Technology and Technical University of Munich. This announcement is especially aimed at future applications such as robots that interact with untreated environments such as logistics, healthcare or elderly medical care. In addition, new interactive online object learning method uses neuromopic computing to support additional object learning even after the robot is deployed.

Intel, along with the Italian Technology Research Institute and Munich University of Technology, demonstrated a new object instance learning using new models on Intel's Neuromopic Research Chip. As a result, while using the power that is about 175 times lower than the existing central processing device (CPU), it has successfully implemented continuous interactive learning with similar or better speed and accuracy to the existing CPU. To this end, the researchers have limited the scope to learn only on a single plastic synapses layer, and implemented a neural network architecture describing other object views on Royrich as needed to secure new neurons as needed. This neural network architecture interacts with the user and was able to develop the learning process autonomously.

The findings are included in the 'Sustainable Interactive Learning for Robots: Neuromopic approach', which was selected as the best papers of the International Neuromopic System (ICON), hosted by the Oak Ridge National Laboratory. have.

Yulia Sandamirskaya, Intel Neuromopic Computing Research Institute Robot Research Officer and the representative of the papers said, When humans learn new objects, they see the details of seeing, returning, and asking objects. Since then, we can immediately recognize the object even in various situations or environments. Our goal is to adapt to unpredictable situations by applying similar functions to future robots that operate in interactive environments, and more naturally with humans To work. The achievements of Roy Hi Chip have further strengthen the value of New Lomo Pick Computing for the future of robotics.

For more information, please visit the official Intel website.