DSpace Community: KAIST School of Electrical Engineering
http://hdl.handle.net/10203/20
KAIST School of Electrical Engineering2024-02-29T10:33:55ZMulticlass autoencoder-based active learning for sensor-based human activity recognition
http://hdl.handle.net/10203/314548
Title: Multiclass autoencoder-based active learning for sensor-based human activity recognition
Authors: Park, Hyunseo; Lee, Gyeong Ho; Han, Jaeseob; Choi, Jun Kyun
Abstract: Leveraging the enormous amounts of real-world data collected through Internet of Things (IoT) technologies, human activity recognition (HAR) has become a crucial component of numerous human-centric applications, with the aim of enhancing the quality of human life. While the recent advancements in deep learning have significantly improved HAR, the process of labeling data continues to remain a significant challenge due to the substantial costs associated with human annotation for supervised model training. Active learning (AL) addresses this issue by strategically selecting informative samples for labeling during model training, thereby enhancing model performance. Although numerous approaches have been proposed for sample selection, which consider aspects of uncertainty and representation, the difficulties in estimating uncertainty and exploiting distribution of high-dimensional data still pose a major issue. Our proposed deep learning-based active learning algorithm, called Multiclass Autoencoder-based Active Learning (MAAL), learns latent representation leveraging the capacity of Deep Support Vector Data Description (Deep SVDD). With the multiclass autoencoder which learns the normal characteristics of each activity class in the latent space, MAAL provides an informative sample selection for model training by establishing a link between the HAR model and the selection model. We evaluate our proposed MAAL using two publicly available datasets. The performance results demonstrate the improvements across the overall active learning rounds, achieving enhancements up to 3.23% accuracy and 3.67% in the F1 score. Furthermore, numerical results and analysis of sample selection are presented to validate the effectiveness of the proposed MAAL compared to the alternative comparison methods.2024-02-01T00:00:00ZIntelligent block copolymer self-assembly towards IoT hardware components
http://hdl.handle.net/10203/318012
Title: Intelligent block copolymer self-assembly towards IoT hardware components
Authors: Yang, Geon Gug; Choi, Hee Jae; Li, Sheng; Kim, Jang Hwan; Kwon, Kyeongha; Jin, Hyeong Min; Kim, Bong Hoon; Kim, Sang Ouk2024-02-01T00:00:00ZOptimal stabilizing rates of switched linear control systems under arbitrary known switchings
http://hdl.handle.net/10203/313558
Title: Optimal stabilizing rates of switched linear control systems under arbitrary known switchings
Authors: Hu, Jianghai; Shen, Jianglai; Lee, Donghwan
Abstract: The problem of stabilizing discrete-time switched linear control systems using continuous control input under arbitrary mode switchings is studied. It is assumed that at each time instance the switching mode can be arbitrarily chosen but is always known by the controller designing the continuous control input; thus the continuous controller is of the general form of an ensemble of mode-dependent state feedback controllers. Under this setting, the fastest (worst-case) stabilizing rate is proposed as a quantitative metric of the systems’ stabilizability. Conditions are derived on when this stabilizing rate can be exactly achieved by an admissible control policy and a counter example is given to show that the stabilizing rate may not always be attained by a mode-dependent linear state feedback control policy. Bounds on the stabilizing rate are derived using (semi)norms. When such bounds are tight, the corresponding extremal norms are characterized geometrically. Numerical algorithms based on ellipsoid and polytope norms are developed for computing bounds of the stabilizing rate and illustrated through examples.2024-01-01T00:00:00ZSn-doped n-type amorphous gallium oxide semiconductor with energy bandgap of 4.9 eV
http://hdl.handle.net/10203/315029
Title: Sn-doped n-type amorphous gallium oxide semiconductor with energy bandgap of 4.9 eV
Authors: Seo, Dahee; Baek, Jongsu; Cho, Byung-Jin; Hwang, Wan Sik
Abstract: An n-type amorphous Ga2O3 thin film with a bandgap of 4.9 eV is formed using the Sn-dopant spin-on-glass (SOG) method followed by a drive-in process at 400 °C for 1 h in an Ar atmosphere. The diffused Sn dopants effectively impact the current enhancement of the amorphous Ga2O3 thin film. The metal-oxide-semiconductor field-effect transistor (MOSFET) with the Sn-doped amorphous Ga2O3 channel layer exhibits typical n-type behavior and well-behaved transistor characteristics presenting two distinct operation regions: linear and saturation. By using a low-temperature deposition method, it becomes possible to create n-type amorphous Ga2O3 thin films with a bandgap of 4.9 eV that are highly compatible with various substrates. Furthermore, ultra-shallow junction formation is achieved in the amorphous Ga2O3 thin film, which is beneficial for high performance device applications.2024-01-01T00:00:00Z