김예성 교수

Human activity recognition is a key task of many Internet of Things (IoT) applications to understand underlying contexts and react with the environments. Machine learning is widely exploited to identify the activities from sensor measurements, however, they are often overcomplex to run on less-powerful IoT devices. In this paper, we present an alternative approach to efficiently support the activity recognition tasks using brain-inspired hyperdimensional (HD) computing. We show how the HD comput

DNA pattern matching is widely applied in many bioinformatics applications. The increasing volume of the DNA data exacerbates the runtime and power consumption to discover DNA patterns. In this paper, we propose a hardware-software co-design, called GenieHD, which efficiently parallelizes the DNA pattern matching task. We exploit brain-inspired hyperdimensional (HD) computing which mimics pattern-based computations in human memory. We transform inherent sequential processes of the DNA pattern ma

The emergence of Internet of Things increases the complexity and the heterogeneity of computing platforms. Migrating workload between various platforms is one way to improve both energy efficiency and performance. Effective migration decisions require accurate estimates of its costs and benefits. To date, these estimates were done by either instrumenting the source code/binaries, thus causing high overhead, or by using power estimates from hardware performance counters, which work well for indiv

In recent years, machine learning for visual object recognition has been applied to various domains, e.g., autonomous vehicle, heath diagnose, and home automation. However, the recognition procedures still consume a lot of processing energy and incur a high cost of data movement for memory accesses. In this paper, we propose a novel hardware accelerator design, called ORCHARD, which processes the object recognition tasks inside memory. The proposed design accelerates both the image feature extra

Mobile devices are severely limited in memory, which affects critical user-experience metrics such as application service time. Emerging non-volatile memory (NVM) technologies such as STT-RAM and PCM are ideal candidates to provide higher memory capacity with negligible energy overhead. However, existing memory management systems overlook mobile users application usage which provides crucial cues for improving user experience. In this paper, we propose CAUSE, a novel memory system based on DRAM-

A low cost, simply structured micropump, which is actuated by piezoelectric discs, was fabricated using polydimethylsiloxane. As a flow-rectifying element, diffusers were used instead of passive check valves. To evaluate the performance of a micropump, the deflection of a diaphragm and the flowrate of a pump were measured experimentally for various applied voltages. The deflection of a glass diaphragm was measured using an atomic force microscope. It increased linearly (up to 0.4 μm) with the vo

Many applications running on high-performance computing systems share limited resources such as the last-level cache, often resulting in lower performance. Intel recently introduced a new control mechanism, called cache allocation technology (CAT), which controls the cache size used by each application. To intelligently utilize this technology for automated management, it is essential to accurately identify application performance behavior for different cache allocation scenarios. In this work,

The brain-inspired hyperdimensional computing (HDC) gains attention as a light-weight and extremely parallelizable learning solution alternative to deep neural networks. Prior research shows the effectiveness of HDC-based learning on less powerful systems such as edge computing devices. However, the many-class classification problem is beyond the focus of mainstream HDC research; the existing HDC would not provide sufficient quality and efficiency due to its coarse-grained training. In this pape

This paper proposes a hardware accelerator design, called object recognition and classification hardware accelerator on resistive devices, which processes object recognition tasks inside emerging nonvolatile memory. The in-memory processing dramatically lowers the overhead of data movement, improving overall system efficiency. The proposed design accelerates key subtasks of image recognition, including text, face, pedestrian, and vehicle recognition. The evaluation shows significant improvements

Behavior of smartphone systems is highly influenced by user interactions, such as `zooming' and `scrolling', which determine the execution phases within applications and lead to different power and performance demands. Current power and thermal management algorithms are agnostic to these behaviors. We propose a novel user activity recognition framework that enables user activity-aware system decisions. The proposed framework carefully monitors system events initiated by user interactions and ide

The purpose of the present study was to examine effects of driver, vehicle, and environment characteristics on Collision Warning System (CWS) design. One hypothesis was made that the capability of collision avoidance would not be same among a driver, vehicle, and environment group with different characteristics. Accident analysis and quantitative analysis was used to examine this hypothesis in terms of ‘risk’ and ‘safety margin’ respectively. Rear-end collision had a stronger focus in the presen

Behavior of smartphone systems is highly influenced by user interactions, such as `zooming' and `scrolling', which determine the execution phases within applications and lead to different power and performance demands. Current power and thermal management algorithms are agnostic to these behaviors. We propose a novel user activity recognition framework that enables user activity-aware system decisions. The proposed framework carefully monitors system events initiated by user interactions and ide

In the previous articles in this series, the problem of calibrating residential location models has been extensively discussed. The discussion has been exclusively devoted to what may be termed trip-end calibration procedures. This was largely the result of the available data sets and the focus on multivariate multiparametric attractiveness functions. Most other authors dealt with what may be called trip-interchange calibration. This article presents a comparison of the two calibration approache

Drawbacks have existed with current pavement crack detection, mapping, and sealing methods which are either solely manual or solely autonomous. A man-machine balance can instead be used, taking advantage of the respective strength of man and machine. This paper describes a man-machine balanced crack sealing process developed for the University of Texas (UT) Automated Road Maintenance Machine (ARMM). Some of the design issues and tradeoffs involved in balancing human and machine functions for aut

Decoherence, often caused by unavoidable coupling with the environment, leads to degradation of quantum coherence. For a multipartite quantum system, decoherence leads to degradation of entanglement and, in certain cases, entanglement sudden death. Tackling decoherence, thus, is a critical issue faced in quantum information, as entanglement is a vital resource for many quantum information applications including quantum computing, quantum cryptography, quantum teleportation, and quantum metrology