How can a hypergraph help to predict the dynamic behavior of complex networked systems?   The study proposes an efficient data-driven model to predict the dynamic behavior of various networked physical systems using a Hypergraph Convolutional Recurrent Neural Network (HGC-RNN). The hypergraph can model the mutual interactions among subcomponents of the target system.   Article | Fall 2020   &nb...

Telecommunication optical fibers can stabilize the laser frequencies to the quadrillionth level   Professor Jungwon Kim’s research team at KAIST has established a simple, compact, alignment-free, and potentially low-cost all-fiber photonic platform for generating multiple ultra-stable optical frequency combs, facilitating various applications ranging from telecommunications through molecular spectroscop...

Rapid analysis of Sr compounds for nuclear forensics   Prof. Sungyeol Choi from the Department of Nuclear and Quantum Engineering (NQe) is developing a rapid analysis method without sample preparations to detect Sr compounds for nuclear forensics and radioactive waste management. Single-shot laser-induced breakdown spectroscopy (LIBS) can detect the thin layers of Sr compounds on metals, porous medium, and t...

SoundWear: How sound-based wearables can facilitate imaginative heads-up play for children   This research indicates how sound augmentation can promote play experiences in physical, social, and imaginative aspects by designing and examining a novel wearable bracelet with which children can explore, select, sonify, and transfer sounds.   Article | Fall 2020     In the past, children used to...

A new analysis algorithm reveals precise 3D spatial organization of the brain   A new analysis technique developed for automated and calibrated neuron mapping into the standardized 3-D mouse brain reference atlas allows a precise analysis of the spatial organization of neural circuits and brings brain structure analysis to a new level.   Article | Fall 2020     A KAIST research team led b...

TensorDIMM: An AI accelerator for personalized recommendation algorithms based on processing-in-memory   TensorDIMM is the first architectural solution tackling sparse embedding layers of personalized recommendation systems. Our solution is based on a practical processing-in-memory architecture that fundamentally addresses the memory capacity and bandwidth challenges of sparse embedding layers widely employe...