187037 research outputs found
Sort by
A pharmaceutical composition for the prevention or treatment of cancer comprising the intestinal microorganisms as an active ingredient
No full text본 발명은 장내 미생물 군집, 특히 암이 발생하지 않은 정상인의 장내 미생물 군집을 이용한 암의 예방 또는 치료용 약학조성물에 관한 것이다. 본 발명의 약학조성물은 암 미발생 개체의 장 내에 존재하는 본래의 공생미생물로서, 화학적항암제 등과 같이 환자의 몸에 부담을 주지 않으면서도 효과적으로 암을 치료하고, 암 환자의 생존율을 향상시키는 효과가 있으므로, 의학 분야에서 크게 이용될 것으로 기대된다
MULTI-TASK CORRUPTED PREDICTION FOR LEARNING ROBUST AUDIO-VISUAL SPEECH REPRESENTATION
No full textAudio-visual speech recognition (AVSR) incorporates auditory and visual modalities to improve recognition accuracy, particularly in noisy environments where audio-only speech systems are insufficient. While previous research has largely addressed audio disruptions, few studies have dealt with visual corruptions, e.g., lip occlusions or blurred videos, which are also detrimental. To address this real-world challenge, we propose CAV2vec, a novel self-supervised speech representation learning framework particularly designed to handle audio-visual joint corruption. CAV2vec employs a self-distillation approach with a corrupted prediction task, where the student model learns to predict clean targets, generated by the teacher model, with corrupted input frames. Specifically, we suggest a unimodal multi-task learning, which distills cross-modal knowledge and aligns the corrupted modalities, by predicting clean audio targets with corrupted videos, and clean video targets with corrupted audios. This strategy mitigates the dispersion in the representation space caused by corrupted modalities, leading to more reliable and robust audiovisual fusion. Our experiments on robust AVSR benchmarks demonstrate that the corrupted representation learning method significantly enhances recognition accuracy across generalized environments involving various types of corruption. Our code is available at https://github.com/sungnyun/cav2vec
MA2E: Addressing Partial Observability in Multi-Agent Reinforcement Learning with Masked Auto-Encoder
No full text
Gas Source Localization in Unknown Indoor Environments Using Dual-Mode Information-Theoretic Search
No full textThis letter proposes a dual-mode planner for localizing gas sources using a mobile sensor in unknown indoor spaces. The complexity of indoor environments creates constraints on search paths, leading to situations where no valid paths can be generated, which are termed as dead end in this letter. The proposed dual-mode planner is designed to effectively address the dead end problem while maintaining efficient search paths. In addition, the absence of analytical dispersion models that can be used in unknown indoor environments presents another critical issue for indoor gas source localization (GSL). To address this, we present an indoor Gaussian dispersion model (IGDM) that can analytically model indoor gas dispersion without a complete map. Finally, we establish a GSL framework for indoor environments along with real-time mapping, utilizing the dual-mode planner and IGDM. This framework is validated in indoor scenarios with the realistic gas dispersion simulator. The simulation results show the high success rate of the proposed method, its ability to reduce search time, and its computational efficiency. Furthermore, through real-world experiments, we demonstrate the potential of the proposed approach as a practical solution, evidenced by its satisfactory performance.
Ternary Transistors With Reconfigurable Polarities
No full textThe recent surge in interest in ultra-power-saving electronic systems has highlighted multi-valued logic circuitry as a promising technology that can simultaneously reduce circuit area, complexity, and power consumption compared to conventional binary logic. Nevertheless, the development of both p-type and n-type multi-valued transistors with stable intermediate states is rare, particularly without CMOS-incompatible heterostructures. Here, polarity-reconfigurable ternary transistors are introduced fabricated using few-layer black phosphorus (BP) homojunction. The ternary transistors feature asymmetric contacts and control gates, which determine carrier polarity and injection levels. The control gates allow the conversion from a conventional ambipolar operation to a p-type ternary operation, with a approximate to 50-fold improvement in the on-off ratio and a well-defined intermediate state. The intermediate state is established by a weakly gate-dependent injection of minority carriers. The operational characteristics are discussed in relation to the applied control gate, drain bias, BP thickness, and contact metal, along with the ternary-to-binary transition. Notably, the devices also exhibit electrical switching to an n-type ternary operation, with its intermediate state matching that of the p-type ternary operation thanks to the antisymmetric device architecture.
Ultrathin 3D Cu/Li Composite with Enhanced Li Utilization for High Energy Density Li-Metal Battery Anodes
No full textLi-metal batteries (LMBs) are promising candidates for next-generation energy storage devices because of their high energy densities. However, limitations of Li-metal anodes (LMAs) such as dendrite formation hinder their practical application. This paper reports an ultrathin 3D Cu/Li composite anode (Li in 3DCu) with a thickness of <30 <mu>m and moderate Li loading of 5 mA h cm(-)(2) via electrochemical etching and electrodeposition, followed by thermal infiltration of Li. The lightweight composite anode achieves a specific capacity of 514 mA h g(-)(1) while effectively reducing current density and suppressing dendritic growth, thus enabling stable performance at high current densities. Novel insights regarding the Li infiltration mechanism are obtained via an integrated analysis of forces, interfacial chemistry, and thermodynamics, offering a comprehensive understanding for Li infiltration. Electrochemical characterization indicates that the proposed composite anode (Li in 3DCu) achieves a 335% improvement in cycle life compared to that when using a conventional anode with Li on a Cu foil in a Li@Cu||LFP cell. This study establishes a robust platform for lightweight high-performance LMAs by combining structural innovations, maximizing Li utilization, and broadens the understanding of infiltration mechanisms to develop next-generation LMBs.
Generation of pulse-position modulated signals having high extinction ratios via an IQ modulator
No full textWe propose and experimentally demonstrate a method for generating M-ary pulse-position modulation (M-PPM) signals having high extinction ratios (ERs). The method utilizes a dual-path structure, wherein a data-driven Mach-Zehnder modulator (MZM) is embedded in one path and a phase shifter in the other. When the two paths are recombined with a relative it-phase shift, the resulting ER can exceed that of the embedded MZM. We implement the proposed structure using a commercially available in-phase and quadrature (IQ) modulator and generate a 100-Mb/s 256-PPM signal. The dynamic ER of the generated signal is measured to be greater than 40 dB. We achieve a receiver sensitivity of 4.49 photons per bit (PPB) at a target BER of 10-3 using the generated 256-PPM signal and an optically pre-amplified receiver. This result is only 2.33 dB above the theoretical limit for optically pre-amplified 256-PPM. (c) 2025 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
Balanced group convolution: an improved group convolution based on approximability estimates
No full textThe performance of neural networks has been significantly improved by increasing the number of channels in convolutional layers. However, this increase in performance comes with a higher computational cost, resulting in numerous studies focused on reducing it. One promising approach to address this issue is group convolution, which effectively reduces the computational cost by grouping channels. However, to the best of our knowledge, there has been no theoretical analysis on how well the group convolution approximates the standard convolution. In this paper, we mathematically analyze the approximation of the group convolution to the standard convolution with respect to the number of groups. Furthermore, we propose a novel variant of the group convolution called balanced group convolution, which shows a higher approximation with a small additional computational cost. We provide experimental results that validate our theoretical findings and demonstrate the superior performance of the balanced group convolution over other variants of group convolution.