Search

Your search keyword '"Wang, Qichao"' showing total 485 results
Start Over Author "Wang, Qichao"
485 results on '"Wang, Qichao"'

1. Recent Advances in Speech Language Models: A Survey

Author

Cui, Wenqian, Yu, Dianzhi, Jiao, Xiaoqi, Meng, Ziqiao, Zhang, Guangyan, Wang, Qichao, Guo, Yiwen, and King, Irwin

Subjects
Computer Science - Computation and Language, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing
Abstract

Large Language Models (LLMs) have recently garnered significant attention, primarily for their capabilities in text-based interactions. However, natural human interaction often relies on speech, necessitating a shift towards voice-based models. A straightforward approach to achieve this involves a pipeline of ``Automatic Speech Recognition (ASR) + LLM + Text-to-Speech (TTS)", where input speech is transcribed to text, processed by an LLM, and then converted back to speech. Despite being straightforward, this method suffers from inherent limitations, such as information loss during modality conversion and error accumulation across the three stages. To address these issues, Speech Language Models (SpeechLMs) -- end-to-end models that generate speech without converting from text -- have emerged as a promising alternative. This survey paper provides the first comprehensive overview of recent methodologies for constructing SpeechLMs, detailing the key components of their architecture and the various training recipes integral to their development. Additionally, we systematically survey the various capabilities of SpeechLMs, categorize the evaluation metrics for SpeechLMs, and discuss the challenges and future research directions in this rapidly evolving field., Comment: Work in progress

Published
2024

2. Parameter-Efficient Fine-Tuning with Discrete Fourier Transform

Author

Gao, Ziqi, Wang, Qichao, Chen, Aochuan, Liu, Zijing, Wu, Bingzhe, Chen, Liang, and Li, Jia

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computation and Language
Abstract

Low-rank adaptation~(LoRA) has recently gained much interest in fine-tuning foundation models. It effectively reduces the number of trainable parameters by incorporating low-rank matrices $A$ and $B$ to represent the weight change, i.e., $\Delta W=BA$. Despite LoRA's progress, it faces storage challenges when handling extensive customization adaptations or larger base models. In this work, we aim to further compress trainable parameters by enjoying the powerful expressiveness of the Fourier transform. Specifically, we introduce FourierFT, which treats $\Delta W$ as a matrix in the spatial domain and learns only a small fraction of its spectral coefficients. With the trained spectral coefficients, we implement the inverse discrete Fourier transform to recover $\Delta W$. Empirically, our FourierFT method shows comparable or better performance with fewer parameters than LoRA on various tasks, including natural language understanding, natural language generation, instruction tuning, and image classification. For example, when performing instruction tuning on the LLaMA2-7B model, FourierFT surpasses LoRA with only 0.064M trainable parameters, compared to LoRA's 33.5M. Our code is released at \url{https://github.com/Chaos96/fourierft}., Comment: Accepted by ICML 2024

Published
2024

6. Language Agents for Detecting Implicit Stereotypes in Text-to-image Models at Scale

Author

Wang, Qichao, Bian, Tian, Yin, Yian, Xu, Tingyang, Cheng, Hong, Meng, Helen M., Zheng, Zibin, Chen, Liang, and Wu, Bingzhe

Subjects
Computer Science - Computers and Society, Computer Science - Computation and Language
Abstract

The recent surge in the research of diffusion models has accelerated the adoption of text-to-image models in various Artificial Intelligence Generated Content (AIGC) commercial products. While these exceptional AIGC products are gaining increasing recognition and sparking enthusiasm among consumers, the questions regarding whether, when, and how these models might unintentionally reinforce existing societal stereotypes remain largely unaddressed. Motivated by recent advancements in language agents, here we introduce a novel agent architecture tailored for stereotype detection in text-to-image models. This versatile agent architecture is capable of accommodating free-form detection tasks and can autonomously invoke various tools to facilitate the entire process, from generating corresponding instructions and images, to detecting stereotypes. We build the stereotype-relevant benchmark based on multiple open-text datasets, and apply this architecture to commercial products and popular open source text-to-image models. We find that these models often display serious stereotypes when it comes to certain prompts about personal characteristics, social cultural context and crime-related aspects. In summary, these empirical findings underscore the pervasive existence of stereotypes across social dimensions, including gender, race, and religion, which not only validate the effectiveness of our proposed approach, but also emphasize the critical necessity of addressing potential ethical risks in the burgeoning realm of AIGC. As AIGC continues its rapid expansion trajectory, with new models and plugins emerging daily in staggering numbers, the challenge lies in the timely detection and mitigation of potential biases within these models.

Published
2023

7. A Machine Learning Method for Predicting Traffic Signal Timing from Probe Vehicle Data

Author

Ugirumurera, Juliette, Severino, Joseph, Bensen, Erik A., Wang, Qichao, and Macfarlane, Jane

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Traffic signals play an important role in transportation by enabling traffic flow management, and ensuring safety at intersections. In addition, knowing the traffic signal phase and timing data can allow optimal vehicle routing for time and energy efficiency, eco-driving, and the accurate simulation of signalized road networks. In this paper, we present a machine learning (ML) method for estimating traffic signal timing information from vehicle probe data. To the authors best knowledge, very few works have presented ML techniques for determining traffic signal timing parameters from vehicle probe data. In this work, we develop an Extreme Gradient Boosting (XGBoost) model to estimate signal cycle lengths and a neural network model to determine the corresponding red times per phase from probe data. The green times are then be derived from the cycle length and red times. Our results show an error of less than 0.56 sec for cycle length, and red times predictions within 7.2 sec error on average.

Published
2023

8. Attention Paper: How Generative AI Reshapes Digital Shadow Industry?

Author

Wang, Qichao, Ma, Huan, Wei, Wentao, Li, Hangyu, Chen, Liang, Zhao, Peilin, Zhao, Binwen, Hu, Bo, Zhang, Shu, Zheng, Zibin, and Wu, Bingzhe

Subjects
Computer Science - Computers and Society, Computer Science - Artificial Intelligence
Abstract

The rapid development of digital economy has led to the emergence of various black and shadow internet industries, which pose potential risks that can be identified and managed through digital risk management (DRM) that uses different techniques such as machine learning and deep learning. The evolution of DRM architecture has been driven by changes in data forms. However, the development of AI-generated content (AIGC) technology, such as ChatGPT and Stable Diffusion, has given black and shadow industries powerful tools to personalize data and generate realistic images and conversations for fraudulent activities. This poses a challenge for DRM systems to control risks from the source of data generation and to respond quickly to the fast-changing risk environment. This paper aims to provide a technical analysis of the challenges and opportunities of AIGC from upstream, midstream, and downstream paths of black/shadow industries and suggest future directions for improving existing risk control systems. The paper will explore the new black and shadow techniques triggered by generative AI technology and provide insights for building the next-generation DRM system.

Published
2023

13. Implications of agricultural success in the Yellow River Basin and its strategy for green development

Author

Gang HE, Zhaohui WANG, Qichao ZHU, Jianbo SHEN, Fusuo ZHANG

Subjects
ecological agriculture, drylands agriculture, irrigation, water-saving technology, policy support, Agriculture (General), S1-972
Abstract

● Ecological fragility and water shortage are key challenges in the Yellow River Basin.● Efficient water use technology in drylands greatly increases crop production.● Water-saving irrigation has been widely adopted and has greatly improved water use.● Changing water use from unregulated and inefficient to intensive and efficient is key solution.● Watershed-scale coordination is a key step towards agriculture green development.The Yellow River Basin is an important food production area and an ecological challenge for China, where environmental protection and water scarcity are the major constraints. For the upper reaches of the Yellow River Basin, optimizing the adoption of chemicals in agricultural production and integrating crops with livestock are the key strategies for protecting the eco-environment. For dryland agriculture in the middle and upper reaches, this study summarizes four aspects of efficient precipitation techniques in terms of collection, storage, conservation, and use, which have greatly improved crop yields and supported dryland crop production. Irrigated agriculture in the middle and lower reaches is the core area of China’s grain production, where the area under water-saving irrigation reached 13.0 Mha in 2018, greatly improving water use. Compared with 1998, cereal production in 2018 increased by 62.2 Mt under similar total water withdrawals (49.7 billion to 51.6 billion m3), and the annual soil erosion at the Tongguan Hydrological Observatory reduced by 584 million m3 in 2018, achieving great success in environmental protection and efficient water use. The Chinese government has set a goal for the Yellow River Basin to become the national leader in environmental protection and efficient water use by 2035. Such a high demand requires the combined efforts of the whole community, as well as the adoption of new technologies, coordinated basin-wide development, and adequate policy support.

Published
2024
Full Text
View/download PDF

22. Prioritized Guidance for Efficient Multi-Agent Reinforcement Learning Exploration

Author

Wang, Qisheng and Wang, Qichao

Subjects
Computer Science - Machine Learning, Computer Science - Multiagent Systems, Statistics - Machine Learning
Abstract

Exploration efficiency is a challenging problem in multi-agent reinforcement learning (MARL), as the policy learned by confederate MARL depends on the collaborative approach among multiple agents. Another important problem is the less informative reward restricts the learning speed of MARL compared with the informative label in supervised learning. In this work, we leverage on a novel communication method to guide MARL to accelerate exploration and propose a predictive network to forecast the reward of current state-action pair and use the guidance learned by the predictive network to modify the reward function. An improved prioritized experience replay is employed to better take advantage of the different knowledge learned by different agents which utilizes Time-difference (TD) error more effectively. Experimental results demonstrates that the proposed algorithm outperforms existing methods in cooperative multi-agent environments. We remark that this algorithm can be extended to supervised learning to speed up its training., Comment: Theequations (7)-(10) in the paper are incorrectly derived, and need to be withdrawn and revised in many places

Published
2019

39. Evolution of Wind Field in the Atmospheric Boundary Layer with using of Multiple Sources Observations during the Transit of Super Typhoon Doksuri (2305).

Author

Wang, Xiaoye, Xu, Jing, Wu, Songhua, Wang, Qichao, Dai, Guangyao, Zhu, Peizhi, Su, Zhizhong, Chen, Sai, Shi, Xiaomeng, and Fan, Mengqi

Subjects
ATMOSPHERIC boundary layer, AUTOMATIC meteorological stations, WIND speed, BOUNDARY layer (Aerodynamics), DOPPLER lidar, TYPHOONS, TROPICAL cyclones
Abstract

The accurate wind field observation of tropical cyclone (TC) boundary layer is of great significance to improve the TC track and intensity forecasting. To investigate the vertical structure of TC boundary layer during the landfall process of Super Typhoon Doksuri, three kinds of instruments including the coherent Doppler lidar (CDL), radar wind profiler (RWP) and automatic weather station (AWS) are deployed at two sites in Xiamen, Fujian province. A data fusion method is developed to obtain the complete wind speed profiles covering the whole Atmospheric Boundary Layer (ABL) based on the above instruments. The wind speeds in the near field blind zones of CDL observation are interpolated by combining the AWS measurements at 10 m. The CDL provides high temporal-spatial resolution wind speed profiles from tens of meters to its highest detection height. The wind speeds above the highest detection height of the CDL would be supplemented with the RWP measurements. The hourly mean wind speed profiles are compared with traditional models. Generally, the wind speed profiles fit well with the power law in the lower part of the ABL, before wind speed changes rapidly. However, it would cause a large error (up to 73 %) to describe the exact wind speed profiles with traditional models during and after the typhoon's passage, especially when the wind speed is almost constant with height or when wind shear exists. Then fine structures and evolutionary processes of the wind field in the ABL during the typhoon landfall are investigated. In addition, the wind field distribution and wind speed variation with distance from the typhoon center are statistical analyzed. The joint wind field measurements of CDL, RWP and AWS have the broad application prospects on the dynamics study of the TC boundary layer and the improvement of the boundary layer parameterization scheme in numerical forecast models. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

40. Simulation Evaluation of a Large-Scale Implementation of Virtual-Phase Link–Based Model Predictive Control.

Author

Shams, Andalib, Wang, Qichao, Ugirumurera, Juliette, Severino, Joseph, Jones, Wesley, and Sanyal, Jibonananda

Subjects
*PREDICTION models, *TRAFFIC signs & signals, *TRAFFIC engineering, *TRAFFIC signal control systems, *TRAFFIC congestion, *ENERGY consumption, *REAL-time control
Abstract

Traffic congestion is a serious problem in the US, and traffic signal control is one of the effective solutions to congestion. Previous research on model predictive control (MPC)-based traffic signal control showed substantial benefits over conventional methods. This study focused on implementing MPC over a large-scale network with complex intersections and the impact of cycle length, network size, and imperfect state estimation on performances. This study implemented a virtual phase link (VPL)-based model predictive control method which used the number of vehicles in each VPL as input state variables and was suitable for National Electrical Manufacturing Association (NEMA) ring-barrier control. To test the impact of network size, the performance of distributed MPC (36 intersections in the network are divided into five subnetworks) was compared with that of MPC over the full network for a set of cycle lengths. To test the impact of imperfect state estimation, we synthetically infused estimation error and developed two scenarios, MPC-error and MPC-error narrow, which had higher and lower estimation errors, respectively. The performance of these MPC methods was compared with that of the existing time-of-day (TOD) method and an offline method that used Webster's method for split and MULTIBAND for cycle length and offset optimization. Trajectory and linkwise signal performance measures were collected from the simulation to evaluate performance. The distributed MPC method with perfect state estimation had the lowest delay and highest energy efficiency of all the methods. The performance of MPC decreased as the prediction inaccuracy increased. MPC-error had 7% and 11% more delay than MPC-error narrow in the morning and evening peaks, respectively. Overall, simulation results suggest that even with imperfect state estimation, MPC methods will outperform offline methods significantly. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

44. Weighted Two-Dimensional Finite Automata

Author

Wang, Qichao, Li, Yongming, Zhou, Wei, Goos, Gerhard, Founding Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Woeginger, Gerhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Du, Ding-Zhu, editor, Li, Lian, editor, Sun, Xiaoming, editor, and Zhang, Jialin, editor

Published
2019
Full Text
View/download PDF

45. A Flexible Cooperative MARL Method for Efficient Passage of an Emergency CAV in Mixed Traffic

Author

Li, Zhi, Wang, Qichao, Wang, Junbo, and He, Zhaocheng

Abstract

Connected and autonomous vehicles offer the possibility to carry out control strategies, thus having great potential to improve traffic efficiency and road safety. The efficient passage of an emergency vehicle calls for the collaborative driving decision-making among multiple vehicles in a dynamically changing local area. However, existing work fails to efficiently adapt to dynamic and complex traffic conditions, thus cannot well solve the task. For better solution, we propose a flexible cooperative multi-agent reinforcement learning approach based on value function factorization, called Q-LSTM. Since the traffic environment is partially observable, the centralized training and decentralized execution paradigm is adopted to learn effective cooperative strategies for individual agents. To flexibly adapt to the changing neighborhood condition around the emergency vehicle, we introduce a long short-term memory network to decompose the learned global value function into local value function of each agent within the neighborhood, whose quantity and entities vary over time. To address the credit assignment problem and realize different roles of the emergency and regular vehicles, reward mechanism and the way agent-wise Q-networks update are well-designed. Extensive experiments are conducted on the Simulation of Urban MObility platform. Results show that our Q-LSTM outperforms state-of-the-art value-based MARL methods. Moreover, the robustness and adaptability of the Q-LSTM are verified in the cases of increased traffic density.

Published
2024
Full Text
View/download PDF

49. Optimization Design and Performance Study of a Heat Exchanger for an Oil and Gas Recovery System in an Oil Depot.

Author

Chen, Zengliang, Luo, Ye, Wang, Zhihui, Liu, Yulin, Gai, Limei, Wang, Qichao, and Hong, Bingyuan

Subjects
HEAT exchangers, PETROLEUM industry, COMPUTATIONAL fluid dynamics, PROCESS capability, CHEMICAL processes
Abstract

High summer temperatures pose numerous challenges to the oil and gas recovery process in oil depots, including reduced adsorption tank recovery rates and decreased absorption tower desorption efficiency. This paper introduces a coupling design approach that integrates chemical process design with computational fluid dynamics simulation. The proposed approach is then utilized to investigate the optimal design and performance of the heat exchanger within the oil depot's oil and gas recovery system. First, according to the given process design parameters, the heat exchanger is preliminary designed to determine the required heat exchange area and heat load. Based on the preliminary design results, a detailed design is carried out, resulting in the following calculations: the hot fluid has inlet and outlet temperatures of 40 °C and 29.52 °C, respectively, with an outlet flow velocity of 9.89 m/s. The cold fluid exhibits inlet and outlet temperatures of 25 °C and 26.98 °C, respectively, with an outlet flow velocity of 0.06 m/s. The specific structure and dimensions of the heat exchanger are determined, including the shell type, pipe specifications, and pipe length. Finally, CFD numerical simulation is utilized to analyze the flow field, velocity field, and pressure field within the designed heat exchanger. The calculations reveal the following findings: the hot fluid exhibited inlet and outlet temperatures of 40 °C and 29.54 °C, respectively, along with an outlet flow velocity of 9.94 m/s. On the other hand, the cold fluid shows inlet and outlet temperatures of 25 °C and 26.39 °C, respectively, with an outlet flow velocity of 0.061 m/s. The results show that the chemical process design and CFD numerical simulation results are consistent and can be mutually verified. The designed heat exchanger can efficiently cool oil and gas from 40 °C to 30 °C, and the oil and gas processing capacity can reach 870 m3/h, which is conducive to realizing the goals of energy saving, environmental protection, and safety. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results