Multi-Agent Deep Reinforcement Learning-Based Resource Allocation in HPC/AI Converged Cluster

As the complexity of deep learning(DL)networks and training data grows enormously,methods that scale with computation are becoming the future of artificial intelligence(AI)development.In this regard,the interplay between machine learning(ML)and high-performance computing(HPC)is an innovative paradigm to speed up the efficiency of AI research and development.However,building and operating an HPC/AI converged system require broad knowledge to leverage the latest computing,networking,and storage technologies.Moreover,an HPC-based AI computing environment needs an appropriate resource allocation and monitoring strategy to efficiently utilize the system resources.In this regard,we introduce a technique for building and operating a high-performance AI-computing environment with the latest technologies.Specifically,an HPC/AI converged system is configured inside Gwangju Institute of Science and Technology(GIST),called GIST AI-X computing cluster,which is built by leveraging the latest Nvidia DGX servers,high-performance storage and networking devices,and various open source tools.Therefore,it can be a good reference for building a small or middlesized HPC/AI converged system for research and educational institutes.In addition,we propose a resource allocation method for DL jobs to efficiently utilize the computing resources with multi-agent deep reinforcement learning(mDRL).Through extensive simulations and experiments,we validate that the proposed mDRL algorithm can help the HPC/AI converged cluster to achieve both system utilization and power consumption improvement.By deploying the proposed resource allocation method to the system,total job completion time is reduced by around 20%and inefficient power consumption is reduced by around 40%.

Programmable directional color dynamics using plasmonics

Adaptive multicolor filters have emerged as key components for ensuring color accuracy and resolution in outdoor visual devices.However,the current state of this technology is still in its infancy and largely reliant on liquid crystal devices that require high voltage and bulky structural designs.Here,we present a multicolor nanofilter consisting of multilayered‘active’plasmonic nanocomposites,wherein metallic nanoparticles are embedded within a conductive polymer nanofilm.These nanocomposites are fabricated with a total thickness below 100 nm using a‘lithography-free’method at the wafer level,and they inherently exhibit three prominent optical modes,accompanying scattering phenomena that produce distinct dichroic reflection and transmission colors.Here,a pivotal achievement is that all these colors are electrically manipulated with an applied external voltage of less than 1 V with 3.5 s of switching speed,encompassing the entire visible spectrum.Furthermore,this electrically programmable multicolor function enables the effective and dynamic modulation of the color temperature of white light across the warm-to-cool spectrum(3250 K-6250 K).This transformative capability is exceptionally valuable for enhancing the performance of outdoor optical devices that are independent of factors such as the sun’s elevation and prevailing weather conditions.

Functional photonic structures for external interaction with flexible/ wearable devices

In addition to vital functions,more subsidiary functions are being expected from wearable devices.The wearable technology thus far has achieved the ability to maintain homeostasis by continuously monitoring physiological signals.The quality of life improves if,through further developments of wearable devices to detect,announce,and even control unperceptive or noxious signals from the environment.Soft materials based on photonic engineering can fulfil the abovementioned functions.Due to the flexibility and zero-power operation of such materials,they can be applied to conventional wearables without affecting existing functions.The achievements to freely tailoring a broad range of electromagnetic waves have encouraged the development of wearable systems for independent recognition/manipulation of light,pollution,chemicals,viruses and heat.Herein,the role that photonic engineering on a flexible platform plays in detecting or reacting to environmental changes is reviewed in terms of material selection,structural design,and regulation mechanisms from the ultraviolet to infrared spectral regions.Moreover,issues emerging with the evolution of the wearable technology,such as Joule heating,battery durability,and user privacy,and the potential solution strategies are discussed.This article provides a systematic review of current progress in wearable devices based on photonic structures as well as an overview of possible ubiquitous advances and their applications,providing diachronic perspectives and future outlook on the rapidly growing research field of wearable technology.