Researchers from Bar-Ilan University have developed a groundbreaking method to drastically reduce the size and energy consumption of deep learning systems—without compromising performance. Published in Physical Review E, their study reveals that by better understanding how deep networks learn, it's possible to prune up to 90% of parameters in certain layers while maintaining accuracy. This advancement, led by Prof. Ido Kanter and PhD student Yarden Tzach, could make AI more efficient, sustainable, and scalable for real-world applications.

Soft electronics, which are designed to function under mechanical deformation (such as bending, stretching, and folding), have become essential in applications like wearable electronics, artificial skin, and brain-machine interfaces. Crystalline silicon is one of the most mature and reliable materials for high-performance electronics; however, its intrinsic brittleness and rigidity pose challenges for integrating it into soft electronics. Recent research has focused on overcoming these limitations by utilizing structural design techniques to impart flexibility and stretchability to Si-based materials, such as transforming them into thin nanomembranes or nanowires. This review summarizes key strategies in geometry engineering for integrating crystalline silicon into soft electronics, from the use of hard silicon islands to creating out-of-plane foldable silicon nanofilms on flexible substrates, and ultimately to shaping silicon nanowires using vapor–liquid–solid or in-plane solid–liquid–solid techniques. We explore the latest developments in Si-based soft electronic devices, with applications in sensors, nanoprobes, robotics, and brain-machine interfaces. Finally, the paper discusses the current challenges in the field and outlines future research directions to enable the widespread adoption of silicon-based flexible electronics.

Scientists have overcome a challenge that has prevented medical ultrasound imaging from being used in intensive care and emergency room settings, an advancement that could someday lead to improved critical care for patients. 

University of Utah engineers encode partial differential equations in light and feed them into newly designed optical neural engine, or ONE, to accelerate machine learning.