Teaching AI to design optical surfaces using real-world imperfections

Designing surfaces that precisely control how light behaves at the nanoscale is tricky. Optical Fourier surfaces, which are ...

Military Inventions That Failed on the Battlefield but Changed Everyday Life

When people think about military innovation, they usually imagine technologies that dominated the battlefield and secured ...
IEEE

Training Lightweight Deep Convolutional Neural Networks Using Bag-of-Features Pooling

Abstract: Convolutional neural networks (CNNs) are predominantly used for several challenging computer vision tasks achieving state-of-the-art performance. However, CNNs are complex models that ...
IEEE

A Novel Small-Data Based Approach for Decoding Yes/No-Decisions of Locked-In Patients Using Generative Adversarial Networks

Abstract: We demonstrate how to use generative adversarial networks to improve the small data problem when training brain-computer-interfaces. The new approach is based on finely graded frequency ...