Connect with us

Artificial Intelligence

Intel Labs Introduces New Approach to Neural Network-Based Object Learning

Updated on
Image: Intel Labs

Researchers at Intel Labs, in collaboration with the Italian Institute of Technology and the Technical University of Munich, have introduced a new approach to neural network-based object learning. The new approach specifically targets future robotics applications like robotic assistants that interact with unconstrained environments, which are present in situations such as logistics and healthcare. 

The new research can prove crucial for improving the service or manufacturing capabilities of our future robots. 

The research paper titled “Interactive continual learning for robots: a neuromorphic approach” was awarded “Best Paper” at the 2022 International Conference on Neuromorphic Systems (ICONS) hosted by Oak Ridge National Laboratory. 

Object Learning and Neuromorphic Computation

New and interactive object learning methods employ neuromorphic computation to enable robots to discover new objects. 

The group of researchers used the new models to demonstrate interactive learning on the Loihi neuromorphic chip, and they achieved up to 175 times lower power consumption when learning new object instances. They also achieved similar or better speed and accuracy when compared to conventional methods run on CPU. 

Image: Intel Labs

The researchers were able to achieve this by implementing a spiking neural network architecture on Loihi, making it possible to localize the learning of the object in a single layer of plastic synapses. It also accounted for different object views by recruiting new neurons on demand. The learning process could then take place autonomously while interacting with the user. 

Yulia Sandamirskaya is senior author of the paper and robotics research lead in Intel’s neuromorphic computing lab.

“When a human learns a new object, they take a look, turn it around, ask what it is, and then they’re able to recognize it again in all kinds of settings and conditions instantaneously,” Sandamirskaya said. “Our goal is to apply similar capabilities to future robots that work in interactive settings, enabling them to adapt to the unforeseen and work more naturally alongside humans. Our results with Loihi reinforce the value of neuromorphic computing for the future of robotics.” 

Image: Intel Labs

Intel Labs Neuromorphic Computing Research

Intel Labs is a leader in the field of neuromorphic computing research, working to “help realize neuromorphic computing’s goal of enabling next-generation intelligent devices and autonomous systems.” 

Neuromorphic computing is guided by the principles of biological neural computation, and it relies on new algorithmic approaches to emulate the human brain and how it interacts with the world.

The innovative architectural approach of neuromorphic computing will be responsible for powering future autonomous AI solutions that require both energy efficiency and continuous learning. It is already being applied in various areas like robotics, sensors, healthcare, and large-scale AI applications.

Alex McFarland is an AI journalist and writer exploring the latest developments in artificial intelligence. He has collaborated with numerous AI startups and publications worldwide.