Search

Your search keyword '"Tian, Zhihua"' showing total 233 results
Start Over Author "Tian, Zhihua"
233 results on '"Tian, Zhihua"'

2. Private Data Valuation and Fair Payment in Data Marketplaces

Author

Tian, Zhihua, Liu, Jian, Li, Jingyu, Cao, Xinle, Jia, Ruoxi, Kong, Jun, Liu, Mengdi, and Ren, Kui

Subjects
Computer Science - Cryptography and Security
Abstract

Data valuation is an essential task in a data marketplace. It aims at fairly compensating data owners for their contribution. There is increasing recognition in the machine learning community that the Shapley value -- a foundational profit-sharing scheme in cooperative game theory -- has major potential to value data, because it uniquely satisfies basic properties for fair credit allocation and has been shown to be able to identify data sources that are useful or harmful to model performance. However, calculating the Shapley value requires accessing original data sources. It still remains an open question how to design a real-world data marketplace that takes advantage of the Shapley value-based data pricing while protecting privacy and allowing fair payments. In this paper, we propose the {\em first} prototype of a data marketplace that values data sources based on the Shapley value in a privacy-preserving manner and at the same time ensures fair payments. Our approach is enabled by a suite of innovations on both algorithm and system design. We firstly propose a Shapley value calculation algorithm that can be efficiently implemented via multiparty computation (MPC) circuits. The key idea is to learn a performance predictor that can directly predict model performance corresponding to an input dataset without performing actual training. We further optimize the MPC circuit design based on the structure of the performance predictor. We further incorporate fair payment into the MPC circuit to guarantee that the data that the buyer pays for is exactly the same as the one that has been valuated. Our experimental results show that the proposed new data valuation algorithm is as effective as the original expensive one. Furthermore, the customized MPC protocol is efficient and scalable., Comment: 14 pages

Published
2022

12. FederBoost: Private Federated Learning for GBDT

Author

Tian, Zhihua, Zhang, Rui, Hou, Xiaoyang, Liu, Jian, and Ren, Kui

Subjects
Computer Science - Cryptography and Security
Abstract

Federated Learning (FL) has been an emerging trend in machine learning and artificial intelligence. It allows multiple participants to collaboratively train a better global model and offers a privacy-aware paradigm for model training since it does not require participants to release their original training data. However, existing FL solutions for vertically partitioned data or decision trees require heavy cryptographic operations. In this paper, we propose a framework named FederBoost for private federated learning of gradient boosting decision trees (GBDT). It supports running GBDT over both vertically and horizontally partitioned data. Vertical FederBoost does not require any cryptographic operation and horizontal FederBoost only requires lightweight secure aggregation. The key observation is that the whole training process of GBDT relies on the ordering of the data instead of the values. We fully implement FederBoost and evaluate its utility and efficiency through extensive experiments performed on three public datasets. Our experimental results show that both vertical and horizontal FederBoost achieve the same level of accuracy with centralized training, where all data are collected in a central server; and they are 4-5 orders of magnitude faster than the state-of-the-art solutions for federated decision tree training; hence offering practical solutions for industrial application., Comment: 15 pages, 8 figures

Published
2020

35. CAFE simulation of directional solidification structure of low nickel austenitic stainless steel

Author

QIN Chen, LI Zhao, ZHAO Liping, JIANG Yajun, CHEN Lichao, ZHANG Huimin, and TIAN Zhihua

Subjects
liquid metal cooling, austenitic stainless steel, numerical simulation, microstructure, Motor vehicles. Aeronautics. Astronautics, TL1-4050
Abstract

In this paper, through the secondary development of ProCAST&CAFE software, the dynamic boundary conditions of solidification process were established, the temperature field and solid fraction in the directional solidification process of low nickel austenitic stainless steel were simulated, and the effect of the pull rate on solidification microstructure was discussed. The results shows that the withdrawal rate is in the range of 15~200 μm/s, as the solidification speed increases, the isotherm becomes dense, gummy zone narrowing. And as the solidification rate increases, the primary dendrite arm gradually refines and the secondary dendrite arm becomes dense and coarse. In addition, as the withdrawal rate increases in the range of 15 to 200 μm/s, the lateral heat dissipation of the edge of the casting is gradually smaller than that of the central longitudinal heat dissipation, and the axial deviation of the dendrites is reduced from 27.66 ° to 25.964 °. At the same time, the comparative analysis simulation results are in good agreement with the experimental results, and the simulation calculation process is more reasonable.

Published
2020
Full Text
View/download PDF

38. Controlling tin whisker growth via oxygen-mediated decomposition of Ti2SnC.

Author

Tang, Haifeng, Yan, Bingzhen, Zhang, Peigen, Yin, Xiaodan, Tian, Zhihua, Das Mahapatra, Susmriti, Zheng, Wei, Tang, Jingwen, and Sun, ZhengMing

Subjects
METALLIC whiskers, WHISKERS, METAL fabrication, TITANIUM, METALS
Abstract

Micro- and nano-sized one-dimensional (1D) metallic materials exhibit tremendous potential in various technological applications such as flexible and transparent electrodes, sensing, and catalysis. However, achieving eco-friendly and large-scale fabrication of diverse 1D metals remains a formidable challenge. The occurrence of A-site whisker growth from MAX phase presents an opportunity for developing an innovative 1D metal fabrication technique. Herein, we investigate the significant role of oxygen in promoting whisker growth from mechanically damaged Ti2SnC MAX phase. Our findings reveal that oxidation of the M-site (titanium) in Ti2SnC phases enhances the formation of atomized tin, which plays a crucial role in tin whisker growth. Furthermore, by employing alternative MAX phase precursors, this technique demonstrates the potential to fabricate 1D metals beyond tin. The diverse A-site elements in MAX phases suggest that this approach could serve as a universal strategy for synthesizing various 1D metallic materials. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

39. Robust evidence reveals the reliable rate of normal/balanced embryos for identifying reciprocal translocation and Robertsonian translocation carriers.

Author

Tian, Zhihua, Lian, Wenchang, Xu, Li, Long, Yanxi, Tang, Li, and Wang, Huawei

Subjects
EMBRYOS, GENETIC counseling, TRANSPORTATION rates, KARYOTYPES, BLASTOCYST
Abstract

Summary: We aimed to evaluate the reliable rate of normal/balanced embryos for reciprocal translocation and Robertsonian translocation carriers and to provide convincing evidence for clinical staff to conduct genetic counselling regarding common structural rearrangements to alleviate patient anxiety. The characteristics of 39,459 embryos that were sourced from unpublished data and literature were analyzed. The samples consisted of 17,536 embryo karyotypes that were not published and 21,923 embryo karyotypes obtained from the literature. Using the PubMed, Cochrane Library, Web of Science, and Embase databases, specific keywords were used to screen the literature for reciprocal translocation and Robertsonian translocation. The ratio of normal/balanced embryos in the overall data was calculated and analyzed, and we grouped the results according to gender to confirm if there were gender differences. We also divided the data into the cleavage stage and blastocyst stage according to the biopsy period to verify if there was a difference in the ratio of normal/balanced embryos. By combining the unpublished data and data derived from the literature, the average rates of normal/balanced embryos for reciprocal translocation and Robertsonian translocation carriers were observed to be 26.96% (7953/29,495) and 41.59% (4144/9964), respectively. Reciprocal translocation and Robertson translocation exhibited higher rates in male carriers than they did in female carriers (49.60% vs. 37.44%; 29.84% vs. 27.67%). Additionally, the data for both translocations exhibited differences in the normal/balanced embryo ratios between the cleavage and blastocyst stages of carriers for both Robertsonian translocation and reciprocal translocation (36.07% vs 43.43%; 24.88% vs 27.67%). The differences between the two location types were statistically significant (P < 0.05). The normal/balanced ratio of embryos in carriers of reciprocal and RobT was higher than the theoretical ratio, and the values ranged from 26.96% to 41.59%. Moreover, the male carriers possessed a higher number of embryos that were normal or balanced. The ratio of normal/balanced embryos in the blastocyst stage was higher than that in the cleavage stage. The results of this study provide a reliable suggestion for future clinic genetic consulting regarding the rate of normal/balanced embryos of reciprocal translocation and Robertsonian translocation carriers. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

48. <inline-formula><tex-math notation="LaTeX">${\sf FederBoost}$</tex-math></inline-formula>: Private Federated Learning for GBDT

Author

Tian, Zhihua, Zhang, Rui, Hou, Xiaoyang, Lyu, Lingjuan, Zhang, Tianyi, Liu, Jian, and Ren, Kui

Abstract

Federated Learning (FL) has been an emerging trend in machine learning and artificial intelligence. It allows multiple participants to collaboratively train a better global model and offers a privacy-aware paradigm for model training since it does not require participants to release their original training data. However, existing FL solutions for vertically partitioned data or decision trees require heavy cryptographic operations. In this article, we propose a framework named $\mathsf {FederBoost}$FederBoost for private federated learning of gradient boosting decision trees (GBDT). It supports running GBDT over both vertically and horizontally partitioned data. Vertical $\mathsf {FederBoost}$FederBoost does not require any cryptographic operation and horizontal $\mathsf {FederBoost}$FederBoost only requires lightweight secure aggregation. The key observation is that the whole training process of GBDT relies on the ordering of the data instead of the values. We fully implement $\mathsf {FederBoost}$FederBoost and evaluate its utility and efficiency through extensive experiments performed on three public datasets. Our experimental results show that both vertical and horizontal $\mathsf {FederBoost}$FederBoost achieve the same level of accuracy with centralized training where all data are collected in a central server; and they are 4-5 orders of magnitude faster than the state-of-the-art solutions for federated decision tree training; hence offering practical solutions for industrial applications.

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results