Search

Your search keyword '"Guo, Bicheng"' showing total 96 results
Start Over Author "Guo, Bicheng"
96 results on '"Guo, Bicheng"'

2. $\alpha$ DARTS Once More: Enhancing Differentiable Architecture Search by Masked Image Modeling

Author

Guo, Bicheng, Guo, Shuxuan, Shi, Miaojing, Chen, Peng, He, Shibo, Chen, Jiming, and Yu, Kaicheng

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Differentiable architecture search (DARTS) has been a mainstream direction in automatic machine learning. Since the discovery that original DARTS will inevitably converge to poor architectures, recent works alleviate this by either designing rule-based architecture selection techniques or incorporating complex regularization techniques, abandoning the simplicity of the original DARTS that selects architectures based on the largest parametric value, namely $\alpha$. Moreover, we find that all the previous attempts only rely on classification labels, hence learning only single modal information and limiting the representation power of the shared network. To this end, we propose to additionally inject semantic information by formulating a patch recovery approach. Specifically, we exploit the recent trending masked image modeling and do not abandon the guidance from the downstream tasks during the search phase. Our method surpasses all previous DARTS variants and achieves state-of-the-art results on CIFAR-10, CIFAR-100, and ImageNet without complex manual-designed strategies.

Published
2022

5. Generalized Global Ranking-Aware Neural Architecture Ranker for Efficient Image Classifier Search

Author

Guo, Bicheng, Chen, Tao, He, Shibo, Liu, Haoyu, Xu, Lilin, Ye, Peng, and Chen, Jiming

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Neural Architecture Search (NAS) is a powerful tool for automating effective image processing DNN designing. The ranking has been advocated to design an efficient performance predictor for NAS. The previous contrastive method solves the ranking problem by comparing pairs of architectures and predicting their relative performance. However, it only focuses on the rankings between two involved architectures and neglects the overall quality distributions of the search space, which may suffer generalization issues. A predictor, namely Neural Architecture Ranker (NAR) which concentrates on the global quality tier of specific architecture, is proposed to tackle such problems caused by the local perspective. The NAR explores the quality tiers of the search space globally and classifies each individual to the tier they belong to according to its global ranking. Thus, the predictor gains the knowledge of the performance distributions of the search space which helps to generalize its ranking ability to the datasets more easily. Meanwhile, the global quality distribution facilitates the search phase by directly sampling candidates according to the statistics of quality tiers, which is free of training a search algorithm, e.g., Reinforcement Learning (RL) or Evolutionary Algorithm (EA), thus it simplifies the NAS pipeline and saves the computational overheads. The proposed NAR achieves better performance than the state-of-the-art methods on two widely used datasets for NAS research. On the vast search space of NAS-Bench-101, the NAR easily finds the architecture with top 0.01$\unicode{x2030}$ performance only by sampling. It also generalizes well to different image datasets of NAS-Bench-201, i.e., CIFAR-10, CIFAR-100, and ImageNet-16-120 by identifying the optimal architectures for each of them.

Published
2022

13. Longitudinal Ultrasonic Vibration-Assisted Planing Method for Processing Micro-Pyramid Arrays.

Author

Gao, Jiashun, Xu, Zhilong, Guo, Bicheng, Lei, Yu, and Yang, Guang

Subjects
OPTICAL elements, SURFACE roughness, PLASTIC films, COPPER, OPTICAL properties, INDUSTRIAL diamonds
Abstract

Micro-pyramid copper molds are critical components in the preparation of high-precision optical elements, such as light-trapping films and reflective films. Their surfaces are composed of micro-pyramid arrays (MPAs). The surface roughness and edge burrs of MPAs seriously affect the optical properties of optical elements. To reduce the surface roughness, as well as the sizes of the edge burrs, the longitudinal ultrasonic vibration-assisted planing (LUVP) method for processing MPAs was developed during this study. In addition, an experiment was conducted to compare the precision planing and LUVP methods of MPA generation. The results show that the tool nose amplitude of the LUVP experimental platform constructed during this study was 3.3 μm, and that the operating frequency was 19.85 kHz. An MPA processed by LUVP had a smaller surface roughness than that of an MPA produced by precision planing; it also had fewer and smaller edge burrs, and there was slightly less diamond tool wear. The MPA cut using the LUVP method had no corrugation on its surface. This research lays a foundation for developing higher-precision micro-pyramid plastic films. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

18. Effect of Carburizing Composite Laser-Shock Processing on Properties and Microstructure of 20CrNiMo Steel.

Author

Chen, Xiuyu, Sun, Jie, Li, Kelin, Lin, Yuru, Xu, Zhilong, Guo, Bicheng, Chen, Junying, and Jiang, Qingshan

Subjects
CARBURIZATION, FUNCTIONALLY gradient materials, MICROSTRUCTURE, RESIDUAL stresses, PEENING, STRESS fractures (Orthopedics)
Abstract

In the service process of gears, premature fatigue failure or fracture of gears is often caused by poor surface performance. 20CrNiMo steel is a commonly used material for gears. Laser-shock peening (LSP), carburizing treatment (CT), and hybrid modification of carburizing treatment and laser-shock peening (LSP + CT) were carried out to improve the performance of 20CrNiMo steel. The hardness, residual stress, microstructure, subgrain size, and toughness of the samples were analyzed following various modification methods. It was observed that the properties of the composite-modified gradient structure materials achieved through carburizing and laser-shock peening were superior to those modified using single methods. After the composite treatment of carburizing and laser shocking, the samples exhibited the most significant increase in hardness, up to 916HV0.1, with a surface layer experiencing residual compressive stress as low as −635 MPa. Simultaneously, a gradient microstructure was formed on the surface layer, with 80% of the crystallites being in the nanoscale range. Furthermore, the toughness was notably enhanced. Experimental results confirm the improvement in the properties of 20CrNiMo samples, resulting in the creation of a functionally graded material through the composite treatment of carburizing and laser shocking. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

21. Latency-Aware Neural Architecture Performance Predictor With Query-to-Tier Technique

Author

Guo, Bicheng, Xu, Lilin, Chen, Tao, Ye, Peng, He, Shibo, Liu, Haoyu, and Chen, Jiming

Abstract

Neural Architecture Search (NAS) is a powerful tool for automating effective image and video processing DNN designing. The ranking of the accuracy has been advocated to design an efficient performance predictor for NAS. The previous contrastive method solves the ranking problem by comparing pairs of architectures and predicting their relative performance. However, it only focuses on the rankings between the two involved architectures and neglects the overall quality distributions of the search space, which may suffer generalization issues. On the contrary, we propose to let the performance predictor concentrate on the global quality level of specific architecture, and learn the tier embeddings of the whole search space automatically with learnable queries. The proposed method, dubbed as Neural Architecture Ranker with Query-to-Tier technique (NARQ2T), explores the quality tiers of the search space globally and classifies each individual to the tier they belong to. Thus, the predictor gains knowledge of the performance distributions of the search space which helps to generalize its ranking ability to the datasets more easily. Thanks to the encoder-decoder design, our method is able to predict the latency of the searched model without deteriorating the performance prediction. Meanwhile, the global quality distribution facilitates the search phase by directly sampling candidates according to the statistics of quality tiers, which is free of training a search algorithm, e.g., Reinforcement Learning or Evolutionary Algorithm, thus it simplifies the NAS pipeline and saves the computational overheads. The proposed NARQ2T achieves state-of-the-art performance on two widely used datasets for NAS research. Moreover, extensive experiments have validated the efficacy of the designed method.

Published
2024
Full Text
View/download PDF

36. Measurement of suction cups release time for press line

Author

Zheng, Xiaowan, Guo, Bicheng, Zhang, Boyang, Fang, Siyuan, Sia, Bernard, and Yang, Lianxiang

Abstract

Suction cups are widely used in press lines to handle and transfer sheet metal. The release time, which directly affects production efficiency and safety, refers to the dropping off period in which the product completely separates from the suction cups. As of now, no release time data is currently available for use in evaluating press line. To provide such data, an experimental setup has been developed. The setup includes vacuum and non-vacuum system designed to simulate the release process of the press line and an electronic triggering system calculating the release time based on controlling the load. The four most commonly used suction cups, namely the large (110 mm diameter) double and single layer suction cups and small (60 mm diameter) double and single layer suction cups, were tested using the setup. The results show that the measurement system is practical and will provide important information for press line optimisation.

Published
2023
Full Text
View/download PDF

47. Reverse micelles synthesis and optical characterization of manganese doped CdSe quantum dots

Author

Guo, Bicheng, Pang, Qi, Yang, Chunlei, Ge, Weikun, Yang, Shihe, Wang, Jiannong, Guo, Bicheng, Pang, Qi, Yang, Chunlei, Ge, Weikun, Yang, Shihe, and Wang, Jiannong

Abstract

CdMnSe Quantum dots were synthesized in reverse micelles with the dot sizes controlled by the water content for the first time. The optical properties of QDs samples with various dot sizes and Mn2+ concentrations were studied both in solution and thin film samples. The electron spin resonance study clarified that both substitutional and surface Mn sites were present. The abnormal temperature behavior of the photoluminescence was believed to originate from the combination of the Mn2+ related emission with the excitonic emission. © 2005 American Institute of Physics.

Published
2005

48. Cd1-xMnxS quantum dots: new synthesis and characterization

Author

Pang, Qi, Guo, Bicheng, Yang, Chunlei, Yang, Shihe, Gong, Menglian, Ge, Weikun, Wang, Jiannong, Pang, Qi, Guo, Bicheng, Yang, Chunlei, Yang, Shihe, Gong, Menglian, Ge, Weikun, and Wang, Jiannong

Abstract

We report on the synthesis of diluted magnetic semiconductor Cd1-xMnxS quantum dots (QDs) by the reverse micelle method using H2S as the sulfur source. The Cd1-xMnxS QDs size ranging from similar to3.2 to similar to4.2nm can be systematically controlled by the amount of H2S added and is characterized by high-resolution transmission electron microscopy. Manganese compositions in the QDs are estimated by energy dispersive spectrometry to be up to similar to3\%. Electron spin resonance spectra analysis indicates that Mn2+ are incorporated into the US lattice by substituting Cd2+ sites. Optical absorption spectra of the Cd1-xMnxS QDs exhibit well-defined excitonic peaks resulting from the quantum-confinement effect. The peak position and absorption edge show systematic changes with the QDs sizes and Mn2+ compositions. (C) 2004 Elsevier B.V. All rights reserved.

Published
2004

49. Chemical synthesis and characterization of CdMnS and CdMnSe quantum dots

Author

Guo, Bicheng and Guo, Bicheng

Abstract

Diluted magnetic semiconductors CdMnS and CdMnSe quantum dots (QDs) are successfully synthesized by reverse micelle method. The QDs are characterized by the high-resolution transmission electron microscopy (HRTEM), energy dispersion spectrometry (EDS), electron spin resonance (ESR), optical absorption and photoluminescence (PL) measurements. For CdMnS QDs, two methods have been employed for the synthesis. One is using Na2S as the sulfur source like what other researchers do, and the other is using H2S as the sulfur source. The latter exhibits a systematic control of the dot size with the amount of H2S incorporated. To our knowledge, this method has so far never been reported by others. HRTEM analysis shows that the CdMnS QDs synthesized with the two methods are both almost spherical in shape and defect free. The quantum confinement effect due to different dot sizes and bowing effect due to different Mn2+ contents are observed in the optical absorption spectra of these QDs. The PL spectra with different dot sizes and Mn2+ concentrations show a broad luminescence band, which is ascribed to emissions associated to Mn2+ and surface trap related states. Similarly, the CdMnSe QDs are synthesized in reverse micelles with the dot sizes controlled by the water content. The optical properties of the QDs samples with various dot sizes and Mn2+ concentrations have been studied both in solution and thin film sample form at room temperature and low temperature, respectively. The temperature dependence and excitation power dependence of the CdMnSe QDs PL spectra show that the luminescence comes from the Mn2+ related emission combined with excitonic emission. For both CdMnS and CdMnSe QDs, the ESR studies identified two Mn2+ sites. One is substituting the Cd2+ site and the other is located near the surface.

Published
2004

Catalog

Books, media, physical & digital resources
See catalog results