Search

Your search keyword '"Qi, Yanxing"' showing total 247 results
Start Over Author "Qi, Yanxing"
247 results on '"Qi, Yanxing"'

17. Reinforcement Learning from Statistical Feedback: the Journey from AB Testing to ANT Testing

Author

Han, Feiyang, Wei, Yimin, Liu, Zhaofeng, Qi, Yanxing, Han, Feiyang, Wei, Yimin, Liu, Zhaofeng, and Qi, Yanxing

Abstract

Reinforcement Learning from Human Feedback (RLHF) has played a crucial role in the success of large models such as ChatGPT. RLHF is a reinforcement learning framework which combines human feedback to improve learning effectiveness and performance. However, obtaining preferences feedback manually is quite expensive in commercial applications. Some statistical commercial indicators are usually more valuable and always ignored in RLHF. There exists a gap between commercial target and model training. In our research, we will attempt to fill this gap with statistical business feedback instead of human feedback, using AB testing which is a well-established statistical method. Reinforcement Learning from Statistical Feedback (RLSF) based on AB testing is proposed. Statistical inference methods are used to obtain preferences for training the reward network, which fine-tunes the pre-trained model in reinforcement learning framework, achieving greater business value. Furthermore, we extend AB testing with double selections at a single time-point to ANT testing with multiple selections at different feedback time points. Moreover, we design numerical experiences to validate the effectiveness of our algorithm framework.

Published
2023

39. Efficient Degradation of Toluene over CeaMnβOX‐X Synthesized by One‐Step Strategy.

Author

Li, Xuelian, Su, Huaigang, Zhang, Chuanwei, Niu, Yongfang, and Qi, Yanxing

Subjects
TOLUENE, BIMETALLIC catalysts, CATALYTIC oxidation, ACTIVATION energy, SURFACE area, CRYSTAL defects
Abstract

Mesoporous Ce−Mn based bimetallic catalysts were prepared by one‐step method using low‐cost precursors under mild conditions. The catalyst prepared at pH=10.5 and Mn/(Ce+Mn)=1/2 showed the optimal performance, and the complete catalytic oxidation of 3000 ppm toluene could be achieved at 215 °C (WHSV=60,000 mL/(g h)). CeMnOX‐10.5 exhibited the highest specific surface area (SBET=151 m2/g), and the normalized reaction rate was 2.013*10−8 mol ⋅ m−2 ⋅ s−1 with the minimum activation energy (Ea=106.6 kJ/mol). Various characterizations confirmed that the synergistic effect between Ce and Mn was the essential reason for promoting the comprehensive performance of catalysts. In addition, the possible structural defects in the catalyst and the schematic diagram of toluene degradation mechanism are clearly provided. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

43. Synthesis of amine-pillar[5]arene porous adsorbent for adsorption of CO2and selectivity over N2and CH4

Author

Li, Hui, Qi, Yanxing, Chen, Jia, Wang, Juanjuan, Yang, Min, and Qiu, Hongdeng

Abstract

A novel amine-modified pillar[5]arene bonded porous silica adsorbent (DETA-P5S) was designed to be applied to dynamic CO2adsorption and selective separation of CO2over N2and CH4gases mixture. The results demonstrated that reasonable introduction of DETA into the BE-P5 bonded silica support has significantly increased the adsorption capacity of CO2. The DETA-P5S has the optimal adsorption capacity of 9.1 mmol/g with 5 vol% CO2at 40 °C. The main reason of this increased capacity could be attributed to the enhanced CO2diffusion into porous adsorbent for its better dispersion in the pores of amine-pillar[5]arene cavity and active site of DETA. Furthermore, the dynamic saturation adsorption capacities of DETA-P5S were 7.11 (0.37) and 6.18 (0.44) mmol/g for CO2/N2and CO2/CH4, respectively, both the gas mixtures showed high separation selectivity. Simultaneously, the DETA-P5S can maintain outstanding CO2adsorption capacity after fifteen regeneration cycles. Consequently, the designed DETA-P5S could serve as a promising adsorbent for CO2capture and storage.

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results