Real-time neural network training on chip

This innovation eliminates the need to transfer trained models from computers, drastically reducing energy consumption and enhancing efficiency.

Researchers at TU Eindhoven have achieved a significant breakthrough in AI technology by developing a neuromorphic chip that supports on-chip neural network training. This innovation eliminates the need to transfer trained models from computers, drastically reducing energy consumption and enhancing efficiency. Utilising electrochemical random-access memory (EC-RAM) components, these chips mimic neuron behaviour, allowing real-time learning and adaptation.

ADVERTISEMENT

Praying mantis inspires machine vision

University of Virginia have developed artificial compound eyes inspired by the praying mantis, addressing critical limitations in self-driving cars and other applications. These biomimetic eyes integrate microlenses and photodiodes within a flexible, hemispherical design, offering enhanced depth perception and a wide view. By mimicking the mantis’ binocular vision it acheives an accurate 3D spatio perception.

Semiconductors turn heat into power

Researchers from Germany, Italy, and the UK have developed a groundbreaking semiconductor by integrating alloys into GeSn semiconductors into the CMOS (Complementary Metal-Oxide-Semiconductor) process, they created a material that can efficiently convert waste heat from computer processors back into electricity.

"Exciting news! Mid-day is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest news!" Click here!