Jian Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jian Wang has a total of 107 co-authors. Here are the top co-authors by number of shared publications:
Jun Wang
Hangzhou Research Center, Huawei Technologies
8 shared publicationsAiping Huang
Institute of Information and Communication Engineering, Zhejiang University, Hangzhou, China
7 shared publicationsRong Li
Hangzhou Research Center, Huawei Technologies
7 shared publicationsWei Wang
Department of Information Science and Electronic Engineering, Zhejiang University and Zhejiang Provincial Key Laboratory of Information Network Technology, Hangzhou, China
6 shared publications- Aiping Huang
College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou, China
5 shared publications Chen Xu
Hangzhou Research Center, Huawei Technologies
4 shared publications
Latest Publications
- Open Access
- 1 July 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the International Conference on Automated Planning and Scheduling
- Vol. 33 (1) , 438-442
Improving Zero-Shot Coordination Performance Based on Policy Similarity
Over these years, multi-agent reinforcement learning have achieved remarkable performance in multi-agent planning and scheduling tasks. It typically follows the self-play setting, where agents are trained by playing with a fixed group of agents. However, in the face of zero-shot coordination, where an agent must coordinate with unseen partners, self-play agents may fail. Several methods have been proposed to handle this problem, but they either take a lot of time or lack generalizability. In this paper, we firstly reveal an important phenomenon: the zero-shot coordination performance is strongly linearly correlated with the similarity between an agent's training partner and testing partner. Inspired by it, we put forward a Similarity-Based Robust Training (SBRT) scheme that improves agents' zero-shot coordination performance by disturbing their partners' actions during training according to a pre-defined policy similarity value. To validate its effectiveness, we apply our scheme to three multi-agent reinforcement learning frameworks and achieve better performance compared with previous methods.
- 1 July 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the International Conference on Automated Planning and Scheduling
- Vol. 33 (1) , 438-442
Open AccessImproving Zero-Shot Coordination Performance Based on Policy Similarity
Over these years, multi-agent reinforcement learning have achieved remarkable performance in multi-agent planning and scheduling tasks. It typically follows the self-play setting, where agents are trained by playing with a fixed group of agents. However, in the face of zero-shot coordination, where an agent must coordinate with unseen partners, self-play agents may fail. Several methods have been proposed to handle this problem, but they either take a lot of time or lack generalizability. In this paper, we firstly reveal an important phenomenon: the zero-shot coordination performance is strongly linearly correlated with the similarity between an agent's training partner and testing partner. Inspired by it, we put forward a Similarity-Based Robust Training (SBRT) scheme that improves agents' zero-shot coordination performance by disturbing their partners' actions during training according to a pre-defined policy similarity value. To validate its effectiveness, we apply our scheme to three multi-agent reinforcement learning frameworks and achieve better performance compared with previous methods.
- 12 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Study on Lithography Defect Reduction for 19nm NAND SADP Process
With the dimension of NAND flash reaching the limit of ArF 193nm immersion scanner by single exposure at half pitch 38nm, the SADP technology (Self Aligned Double Patterning) is utilized to achieve 19nm half pitch CD. However, with the line CD shrinking and cell density increasing, the line broken defects and block etch defects, which are resulting from lithography processing, come to be the key yield killer definitely. In this paper, the defect reduction methodology of HLMC 19nm NAND SADP lithography process, including the line broken defects and block etch defects, is summarized by theoretical analysis and tested by experiment results. The defects are eliminated by RRC pre-clean and enhanced rinse after development, which helps to realize 19NAND mass production and yield improvement, and benefits to more advanced technology nodes development as well.
- 12 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Study on Lithography Defect Reduction for 19nm NAND SADP Process
With the dimension of NAND flash reaching the limit of ArF 193nm immersion scanner by single exposure at half pitch 38nm, the SADP technology (Self Aligned Double Patterning) is utilized to achieve 19nm half pitch CD. However, with the line CD shrinking and cell density increasing, the line broken defects and block etch defects, which are resulting from lithography processing, come to be the key yield killer definitely. In this paper, the defect reduction methodology of HLMC 19nm NAND SADP lithography process, including the line broken defects and block etch defects, is summarized by theoretical analysis and tested by experiment results. The defects are eliminated by RRC pre-clean and enhanced rinse after development, which helps to realize 19NAND mass production and yield improvement, and benefits to more advanced technology nodes development as well.
- 9 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 649-652
Co-Attentive Span Network with Multi-task learning for Biomedical Named Entity Recognition
Biomedical Named Entity Recognition (BioNER) is often modeled as a sequence labeling task, which assigns the predefined label to each token in given input sequence. Although these sequential labeling models achieve significant achievements, they often fail to give precise boundaries of the named entity. In addition, a vast amount of work focuses more on textual sequence representation but ignores label information. To tackle these problems, in this paper, we directly model span-level named entity recognition, specifically, we treat the BioNER as a joint task of boundary detection and span classification under a multitask framework. In order to enhance boundary supervision, we introduce an entity type label as an additional guide and propose a co-attentive interactive mechanism to improve the span representation. Extensive experiments 1 on four benchmark datasets demonstrate that our proposed method obtains competitive results, achieving 90.26%, 78.04%, 90.21%, and 86.58% on BC5CDR, JNLPBA, NCBI, and BC2GM datasets, respectively, in terms of F1 score.
- 9 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 649-652
Co-Attentive Span Network with Multi-task learning for Biomedical Named Entity Recognition
Biomedical Named Entity Recognition (BioNER) is often modeled as a sequence labeling task, which assigns the predefined label to each token in given input sequence. Although these sequential labeling models achieve significant achievements, they often fail to give precise boundaries of the named entity. In addition, a vast amount of work focuses more on textual sequence representation but ignores label information. To tackle these problems, in this paper, we directly model span-level named entity recognition, specifically, we treat the BioNER as a joint task of boundary detection and span classification under a multitask framework. In order to enhance boundary supervision, we introduce an entity type label as an additional guide and propose a co-attentive interactive mechanism to improve the span representation. Extensive experiments 1 on four benchmark datasets demonstrate that our proposed method obtains competitive results, achieving 90.26%, 78.04%, 90.21%, and 86.58% on BC5CDR, JNLPBA, NCBI, and BC2GM datasets, respectively, in terms of F1 score.
- Open Access
Smart Scheduling based on Deep Reinforcement Learning for Cellular Networks
To improve the system performance towards the Shannon limit, advanced radio resource management mechanisms play a fundamental role. In particular, scheduling should receive much attention, because it allocates radio resources among different users in terms of their channel conditions and QoS requirements. The difficulties of scheduling algorithms are the tradeoffs need to be made among multiple objectives, such as throughput, fairness and packet drop rate. We propose a smart scheduling scheme based on deep reinforcement learning (DRL). We not only verify the performance gain achieved, but also provide implementation-friend designs, i.e., a scalable neural network design for the agent and a virtual environment training framework. With the scalable neural network design, the DRL agent can easily handle the cases when the number of active users is time-varying without the need to redesign and retrain the DRL agent. Training the DRL agent in a virtual environment offline first and using it as the initial version in the practical usage helps to prevent the system from suffering from performance and robustness degradation due to the time-consuming training. Through both simulations and field tests, we show that the DRL-based smart scheduling outperforms the conventional scheduling method and can be adopted in practical systems.
Open AccessSmart Scheduling based on Deep Reinforcement Learning for Cellular Networks
To improve the system performance towards the Shannon limit, advanced radio resource management mechanisms play a fundamental role. In particular, scheduling should receive much attention, because it allocates radio resources among different users in terms of their channel conditions and QoS requirements. The difficulties of scheduling algorithms are the tradeoffs need to be made among multiple objectives, such as throughput, fairness and packet drop rate. We propose a smart scheduling scheme based on deep reinforcement learning (DRL). We not only verify the performance gain achieved, but also provide implementation-friend designs, i.e., a scalable neural network design for the agent and a virtual environment training framework. With the scalable neural network design, the DRL agent can easily handle the cases when the number of active users is time-varying without the need to redesign and retrain the DRL agent. Training the DRL agent in a virtual environment offline first and using it as the initial version in the practical usage helps to prevent the system from suffering from performance and robustness degradation due to the time-consuming training. Through both simulations and field tests, we show that the DRL-based smart scheduling outperforms the conventional scheduling method and can be adopted in practical systems.
- Open Access
- 1 March 2021
- Published by IOP Publishing in IOP Conference Series: Earth and Environmental Science
- Vol. 702 (1) , 012018
Construction of wind turbine performance model based on SCADA data
Compared with traditional wind turbine performance evaluation methods such as operating power curve fitting, the wind turbine performance model construction method using SCADA data on machine learning has gradually attracted attention in the industry in recent years. In this technical route, the choice of input SCADA data type will directly affect the accuracy of the wind turbine performance model construction. However, in the past, selecting key SCADA data types in the performance evaluation of wind turbines were mostly subjective judgments or correlation analyses. This paper proposed a key SCADA data type selection method based on mutual information calculation and a wind turbine performance model construction method based on deep neural network training. The selected key SCADA data types based on mutual information were applied to the deep neural network to construct the wind turbine performance model, and the actual SCADA operation data of the wind farm was applied to test the model. The results show that the model construction method’s accuracy and generalizability using the above technical route can satisfy the industry demand in the same wind farm with the same wind turbine type.
- 1 March 2021
- Published by IOP Publishing in IOP Conference Series: Earth and Environmental Science
- Vol. 702 (1) , 012018
Open AccessConstruction of wind turbine performance model based on SCADA data
Compared with traditional wind turbine performance evaluation methods such as operating power curve fitting, the wind turbine performance model construction method using SCADA data on machine learning has gradually attracted attention in the industry in recent years. In this technical route, the choice of input SCADA data type will directly affect the accuracy of the wind turbine performance model construction. However, in the past, selecting key SCADA data types in the performance evaluation of wind turbines were mostly subjective judgments or correlation analyses. This paper proposed a key SCADA data type selection method based on mutual information calculation and a wind turbine performance model construction method based on deep neural network training. The selected key SCADA data types based on mutual information were applied to the deep neural network to construct the wind turbine performance model, and the actual SCADA operation data of the wind farm was applied to test the model. The results show that the model construction method’s accuracy and generalizability using the above technical route can satisfy the industry demand in the same wind farm with the same wind turbine type.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Realistic Channel Models Pre-Training
In this paper, we propose a neural-network-based realistic channel model with both the similar accuracy as deterministic channel models and uniformity as stochastic channel models. To facilitate this realistic channel modeling, a multi-domain channel embedding method combined with self-attention mechanism is proposed to extract channel features from multiple domains simultaneously. This "one model to fit them all" solution employs available wireless channel data as the only data set for self-supervised pre-training. With the permission of users, network operators or other organizations can make use of some available user specific data to fine-tune this pre-trained realistic channel model for applications on channel-related downstream tasks. Moreover, even without fine-tuning, we show that the pre-trained realistic channel model itself is a great tool with its understanding of wireless channel.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Realistic Channel Models Pre-Training
In this paper, we propose a neural-network-based realistic channel model with both the similar accuracy as deterministic channel models and uniformity as stochastic channel models. To facilitate this realistic channel modeling, a multi-domain channel embedding method combined with self-attention mechanism is proposed to extract channel features from multiple domains simultaneously. This "one model to fit them all" solution employs available wireless channel data as the only data set for self-supervised pre-training. With the permission of users, network operators or other organizations can make use of some available user specific data to fine-tune this pre-trained realistic channel model for applications on channel-related downstream tasks. Moreover, even without fine-tuning, we show that the pre-trained realistic channel model itself is a great tool with its understanding of wireless channel.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2689-2694
Leader-Following Output Formation Control of Heterogeneous Multi-Agent Systems
A leader-following output formation control problem of heterogeneous multi-agent systems is studied in this work. The system structure of the autonomous leader is different with that of the formation followers. Only the agents which have the leader agent as its neighbour can obtain the leader information. An output formation protocol is constructed to solve the output formation problem based on the output regulation theory. The leader information is necessary for the output formation protocol. Thus the leader information is reconstructed for each formation followers by a leader mission trajectory observer using only the distributed information from the information transmission networks. The output formation describes the scenario that the output trajectories of the formation followers form a special geometrical shape in according with the moving of output trajectories of the autonomous leader. A simulation example shows the effectiveness of the output formation results.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2689-2694
Leader-Following Output Formation Control of Heterogeneous Multi-Agent Systems
A leader-following output formation control problem of heterogeneous multi-agent systems is studied in this work. The system structure of the autonomous leader is different with that of the formation followers. Only the agents which have the leader agent as its neighbour can obtain the leader information. An output formation protocol is constructed to solve the output formation problem based on the output regulation theory. The leader information is necessary for the output formation protocol. Thus the leader information is reconstructed for each formation followers by a leader mission trajectory observer using only the distributed information from the information transmission networks. The output formation describes the scenario that the output trajectories of the formation followers form a special geometrical shape in according with the moving of output trajectories of the autonomous leader. A simulation example shows the effectiveness of the output formation results.
- 1 October 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Deep Reinforcement Learning for Scheduling in Cellular Networks
Integrating artificial intelligence (AI) into wireless networks has drawn significant interest in both industry and academia. A common solution is to replace partial or even all modules in the conventional systems, which is often lack of efficiency and robustness due to their ignoring of expert knowledge. In this paper, we take deep reinforcement learning (DRL) based scheduling as an example to investigate how expert knowledge can help with AI module in cellular networks. A simulation platform, which has considered link adaption, feedback and other practical mechanisms, is developed to facilitate the investigation. Besides the traditional way, which is learning directly from the environment, for training DRL agent, we propose two novel methods, i.e., learning from a dual AI module and learning from the expert solution. The results show that, for the considering scheduling problem, DRL training procedure can be improved on both performance and convergence speed by involving the expert knowledge. Hence, instead of replacing conventional scheduling module in the system, adding a newly introduced AI module, which is capable to interact with the conventional module and provide more flexibility, is a more feasible solution.
- 1 October 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Deep Reinforcement Learning for Scheduling in Cellular Networks
Integrating artificial intelligence (AI) into wireless networks has drawn significant interest in both industry and academia. A common solution is to replace partial or even all modules in the conventional systems, which is often lack of efficiency and robustness due to their ignoring of expert knowledge. In this paper, we take deep reinforcement learning (DRL) based scheduling as an example to investigate how expert knowledge can help with AI module in cellular networks. A simulation platform, which has considered link adaption, feedback and other practical mechanisms, is developed to facilitate the investigation. Besides the traditional way, which is learning directly from the environment, for training DRL agent, we propose two novel methods, i.e., learning from a dual AI module and learning from the expert solution. The results show that, for the considering scheduling problem, DRL training procedure can be improved on both performance and convergence speed by involving the expert knowledge. Hence, instead of replacing conventional scheduling module in the system, adding a newly introduced AI module, which is capable to interact with the conventional module and provide more flexibility, is a more feasible solution.
- 1 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Predicting the Mumble of Wireless Channel with Sequence-to-Sequence Models
Accurate prediction of fading channel in the upcoming transmission frame is essential to realize adaptive transmission for transmitters, and receivers with the ability of channel prediction can also save some computations of channel estimation. However, due to the rapid channel variation and channel estimation error, reliable prediction is hard to realize. In this situation, an appropriate channel model should be selected, which can cover both the statistical model and small scale fading of channel, this reminds us the natural languages, which also have statistical word frequency and specific sentences. Accordingly, in this paper, we take wireless channel model as a language model, and the time-varying channel as talking in this language, while the realistic noisy estimated channel can be compared with mumbling. Furthermore, in order to utilize as much as possible the information a channel coefficient takes, we discard the conventional two features of absolute value and phase, replacing with hundreds of features which will be learned by our channel model, to do this, we use a vocabulary to map a complex channel coefficient into an ID, which is represented by a vector of real numbers. Recurrent neural networks technique is used as its good balance between memorization and generalization, moreover, we creatively introduce sequence-to-sequence (seq2seq) models in time series channel prediction, which can translate past channel into future channel. The results show that realistic channel prediction with superior performance relative to channel estimation is attainable.
- 1 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Predicting the Mumble of Wireless Channel with Sequence-to-Sequence Models
Accurate prediction of fading channel in the upcoming transmission frame is essential to realize adaptive transmission for transmitters, and receivers with the ability of channel prediction can also save some computations of channel estimation. However, due to the rapid channel variation and channel estimation error, reliable prediction is hard to realize. In this situation, an appropriate channel model should be selected, which can cover both the statistical model and small scale fading of channel, this reminds us the natural languages, which also have statistical word frequency and specific sentences. Accordingly, in this paper, we take wireless channel model as a language model, and the time-varying channel as talking in this language, while the realistic noisy estimated channel can be compared with mumbling. Furthermore, in order to utilize as much as possible the information a channel coefficient takes, we discard the conventional two features of absolute value and phase, replacing with hundreds of features which will be learned by our channel model, to do this, we use a vocabulary to map a complex channel coefficient into an ID, which is represented by a vector of real numbers. Recurrent neural networks technique is used as its good balance between memorization and generalization, moreover, we creatively introduce sequence-to-sequence (seq2seq) models in time series channel prediction, which can translate past channel into future channel. The results show that realistic channel prediction with superior performance relative to channel estimation is attainable.
- 1 December 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
HMNPPID: A Database of Protein-protein Interactions Associated with Human Malignant Neoplasms
Protein-protein interactions (PPIs) play a key role in varies aspects of the structural and functional organization of the cell. Knowledge about them unveils the molecular mechanisms of biological processes. At the same time, the PPIs related to human malignant neoplasms can shed light on the biology behind these neoplasms. However, there is no such PPI database available so far. We constructed a database of protein-protein interactions associated with human malignant neoplasms, named HMNPPID, which contains the PPIs of 171 kinds of human malignant neoplasms. HMNPPID provides healthcare professionals with the PPIs of specific malignant neoplasm without digging into amounts of biomedical literatures, which significantly improves the efficiency of researches on the PPIs of malignant neoplasms. In addition, a visualization program, VisualPPI, is provided to facilitate the analysis of the specific PPI network for a malignant neoplasm.
- 1 December 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
HMNPPID: A Database of Protein-protein Interactions Associated with Human Malignant Neoplasms
Protein-protein interactions (PPIs) play a key role in varies aspects of the structural and functional organization of the cell. Knowledge about them unveils the molecular mechanisms of biological processes. At the same time, the PPIs related to human malignant neoplasms can shed light on the biology behind these neoplasms. However, there is no such PPI database available so far. We constructed a database of protein-protein interactions associated with human malignant neoplasms, named HMNPPID, which contains the PPIs of 171 kinds of human malignant neoplasms. HMNPPID provides healthcare professionals with the PPIs of specific malignant neoplasm without digging into amounts of biomedical literatures, which significantly improves the efficiency of researches on the PPIs of malignant neoplasms. In addition, a visualization program, VisualPPI, is provided to facilitate the analysis of the specific PPI network for a malignant neoplasm.