Bingqing Liu
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
Bingqing Liu has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Xikun Huang
Chinese Academy of Sciences
2 shared publications - Ivan Kiskin
University of Oxford
1 shared publication - Stephen Roberts
University of Oxford
1 shared publication
Latest Publications
- 24 June 2024
- Published by Springer Nature
- p. 127-142
A Case-Based Reasoning and Explaining Model for Temporal Point Process
- 24 June 2024
- Published by Springer Nature
- p. 127-142
A Case-Based Reasoning and Explaining Model for Temporal Point Process
- 17 June 2024
- Published by Springer Nature in Data Mining and Knowledge Discovery
- Vol. 38 (6) , 3449-3472
Modelling event sequence data by type-wise neural point process
- 17 June 2024
- Published by Springer Nature in Data Mining and Knowledge Discovery
- Vol. 38 (6) , 3449-3472
Modelling event sequence data by type-wise neural point process
- Open Access
Cumulative Hazard Function Based Efficient Multivariate Temporal Point Process Learning
Most existing temporal point process models are characterized by conditional intensity function. These models often require numerical approximation methods for likelihood evaluation, which potentially hurts their performance. By directly modelling the integral of the intensity function, i.e., the cumulative hazard function (CHF), the likelihood can be evaluated accurately, making it a promising approach. However, existing CHF-based methods are not well-defined, i.e., the mathematical constraints of CHF are not completely satisfied, leading to untrustworthy results. For multivariate temporal point process, most existing methods model intensity (or density, etc.) functions for each variate, limiting the scalability. In this paper, we explore using neural networks to model a flexible but well-defined CHF and learning the multivariate temporal point process with low parameter complexity. Experimental results on six datasets show that the proposed model achieves the state-of-the-art performance on data fitting and event prediction tasks while having significantly fewer parameters and memory usage than the strong competitors. The source code and data can be obtained from https://github.com/lbq8942/NPP.
Open AccessCumulative Hazard Function Based Efficient Multivariate Temporal Point Process Learning
Most existing temporal point process models are characterized by conditional intensity function. These models often require numerical approximation methods for likelihood evaluation, which potentially hurts their performance. By directly modelling the integral of the intensity function, i.e., the cumulative hazard function (CHF), the likelihood can be evaluated accurately, making it a promising approach. However, existing CHF-based methods are not well-defined, i.e., the mathematical constraints of CHF are not completely satisfied, leading to untrustworthy results. For multivariate temporal point process, most existing methods model intensity (or density, etc.) functions for each variate, limiting the scalability. In this paper, we explore using neural networks to model a flexible but well-defined CHF and learning the multivariate temporal point process with low parameter complexity. Experimental results on six datasets show that the proposed model achieves the state-of-the-art performance on data fitting and event prediction tasks while having significantly fewer parameters and memory usage than the strong competitors. The source code and data can be obtained from https://github.com/lbq8942/NPP.
- Open Access
Interpretable Neural Temporal Point Processes for Modelling Electronic Health Records
Electronic Health Records (EHR) can be represented as temporal sequences that record the events (medical visits) from patients. Neural temporal point process (NTPP) has achieved great success in modeling event sequences that occur in continuous time space. However, due to the black-box nature of neural networks, existing NTPP models fall short in explaining the dependencies between different event types. In this paper, inspired by word2vec and Hawkes process, we propose an interpretable framework inf2vec for event sequence modelling, where the event influences are directly parameterized and can be learned end-to-end. In the experiment, we demonstrate the superiority of our model on event prediction as well as type-type influences learning.
Open AccessInterpretable Neural Temporal Point Processes for Modelling Electronic Health Records
Electronic Health Records (EHR) can be represented as temporal sequences that record the events (medical visits) from patients. Neural temporal point process (NTPP) has achieved great success in modeling event sequences that occur in continuous time space. However, due to the black-box nature of neural networks, existing NTPP models fall short in explaining the dependencies between different event types. In this paper, inspired by word2vec and Hawkes process, we propose an interpretable framework inf2vec for event sequence modelling, where the event influences are directly parameterized and can be learned end-to-end. In the experiment, we demonstrate the superiority of our model on event prediction as well as type-type influences learning.
- Open Access
Link-aware link prediction over temporal graph by pattern recognition
A temporal graph can be considered as a stream of links, each of which represents an interaction between two nodes at a certain time. On temporal graphs, link prediction is a common task, which aims to answer whether the query link is true or not. To do this task, previous methods usually focus on the learning of representations of the two nodes in the query link. We point out that the learned representation by their models may encode too much information with side effects for link prediction because they have not utilized the information of the query link, i.e., they are link-unaware. Based on this observation, we propose a link-aware model: historical links and the query link are input together into the following model layers to distinguish whether this input implies a reasonable pattern that ends with the query link. During this process, we focus on the modeling of link evolution patterns rather than node representations. Experiments on six datasets show that our model achieves strong performances compared with state-of-the-art baselines, and the results of link prediction are interpretable. The code and datasets are available on the project website: https://github.com/lbq8942/TGACN.
Open AccessLink-aware link prediction over temporal graph by pattern recognition
A temporal graph can be considered as a stream of links, each of which represents an interaction between two nodes at a certain time. On temporal graphs, link prediction is a common task, which aims to answer whether the query link is true or not. To do this task, previous methods usually focus on the learning of representations of the two nodes in the query link. We point out that the learned representation by their models may encode too much information with side effects for link prediction because they have not utilized the information of the query link, i.e., they are link-unaware. Based on this observation, we propose a link-aware model: historical links and the query link are input together into the following model layers to distinguish whether this input implies a reasonable pattern that ends with the query link. During this process, we focus on the modeling of link evolution patterns rather than node representations. Experiments on six datasets show that our model achieves strong performances compared with state-of-the-art baselines, and the results of link prediction are interpretable. The code and datasets are available on the project website: https://github.com/lbq8942/TGACN.
- 19 July 2023
- Published by Springer Nature
- p. 325-337
Link-Aware Link Prediction over Temporal Graph by Pattern Recognition
- 19 July 2023
- Published by Springer Nature
- p. 325-337
Link-Aware Link Prediction over Temporal Graph by Pattern Recognition
- Open Access
- 10 June 2022
- Published by Louisiana State University Libraries
Phytoplankton and Carbon Dynamics in the Estuarine-Coastal Waters of the Northern Gulf of Mexico from Field Data and Ocean Color Remote Sensing
In this study, phytoplankton community and carbon dynamics were examined in the optically complex estuarine-coastal regions of the northern Gulf of Mexico (nGOM) from field and satellite ocean color observations. As part of this study, bio-optical ocean color algorithms for i) dissolved organic carbon (DOC), ii) phytoplankton pigment composition, iii) adaptive estimation of Chl a and iv) phytoplankton size fractions were developed to facilitate the study of biogeochemical cycling in the nGOM. The phytoplankton based algorithms were applied to Sentinel 3A/B-OLCI oean color data to assess phytoplankton community dynamics to extreme river discharge conditions as well as hurricanes in the nGOM. This study revealed that the effects of hurricanes on phytoplankton community dynamics were dependent on background nutrient conditions, as well as the intensity, track and translational speed of storms: 1) Strong flooding associated with Hurricane Harvey (2017) shifted the dominance of phytoplankton community in Galveston Bay from cyanobacteria and dinoflagellate to diatom and chlorophyte; 2) high levels of organic matter delivered from estuaries to shelf waters after Hurricane Michael (2018) fueled a red tide mixed with coccolithophore bloom in the nGoM; 3) the physical and chemical environment after hurricanes are favorable for the growth and dominance of coccolithophores in shelf waters. Further, microphytoplankton mainly controlled by freshwater inflows showed dominance in estuaries of the nGoM, with highest/lowest values observed in spring/fall. In comparison, phytoplankton size fraction (PSF) dynamics in the midshelf and offshore waters of the nGoM are strongly influenced by Loop Current (LC) expansion, and eddy shedding with highest picophytoplankton fraction observed in the warm waters of LC. DOC dynamics was studied using an empirical algorithm that was developed and applied to multiple satellite sensors (Landsat 5 TM and MODIS-Aqua) to assess multi-decadal (1985-2012) DOC trends in Barataria Basin. The linkages between DOC and environmenal variations were investigated. The relationships between satellite-derived DOC and land cover variations (1985–2011) derived from Landsat-5 TM supervised classification indicate soil loss in the salt marsh to be an important DOC source in the wetland-estuary system, and overall strong land use/land loss impact on the long-term DOC trends in the Barataria Basin.
- 10 June 2022
- Published by Louisiana State University Libraries
Open AccessPhytoplankton and Carbon Dynamics in the Estuarine-Coastal Waters of the Northern Gulf of Mexico from Field Data and Ocean Color Remote Sensing
In this study, phytoplankton community and carbon dynamics were examined in the optically complex estuarine-coastal regions of the northern Gulf of Mexico (nGOM) from field and satellite ocean color observations. As part of this study, bio-optical ocean color algorithms for i) dissolved organic carbon (DOC), ii) phytoplankton pigment composition, iii) adaptive estimation of Chl a and iv) phytoplankton size fractions were developed to facilitate the study of biogeochemical cycling in the nGOM. The phytoplankton based algorithms were applied to Sentinel 3A/B-OLCI oean color data to assess phytoplankton community dynamics to extreme river discharge conditions as well as hurricanes in the nGOM. This study revealed that the effects of hurricanes on phytoplankton community dynamics were dependent on background nutrient conditions, as well as the intensity, track and translational speed of storms: 1) Strong flooding associated with Hurricane Harvey (2017) shifted the dominance of phytoplankton community in Galveston Bay from cyanobacteria and dinoflagellate to diatom and chlorophyte; 2) high levels of organic matter delivered from estuaries to shelf waters after Hurricane Michael (2018) fueled a red tide mixed with coccolithophore bloom in the nGoM; 3) the physical and chemical environment after hurricanes are favorable for the growth and dominance of coccolithophores in shelf waters. Further, microphytoplankton mainly controlled by freshwater inflows showed dominance in estuaries of the nGoM, with highest/lowest values observed in spring/fall. In comparison, phytoplankton size fraction (PSF) dynamics in the midshelf and offshore waters of the nGoM are strongly influenced by Loop Current (LC) expansion, and eddy shedding with highest picophytoplankton fraction observed in the warm waters of LC. DOC dynamics was studied using an empirical algorithm that was developed and applied to multiple satellite sensors (Landsat 5 TM and MODIS-Aqua) to assess multi-decadal (1985-2012) DOC trends in Barataria Basin. The linkages between DOC and environmenal variations were investigated. The relationships between satellite-derived DOC and land cover variations (1985–2011) derived from Landsat-5 TM supervised classification indicate soil loss in the salt marsh to be an important DOC source in the wetland-estuary system, and overall strong land use/land loss impact on the long-term DOC trends in the Barataria Basin.
- 1 February 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2, 681-686
An Overview of Gaussian process Regression for Volatility Forecasting
Forecasting financial time series using trading data has held the attention of academics and practitioners due to the complexity of the financial system and the profit it can generate for investors. Although investors aim to achieve a consistent attainment of returns, it is important for investors to understand the concept and measurement of volatility. A higher volatility indicates a wider potential range of future returns. But with higher potential returns, comes a higher potential risk. With Gaussian Processes (GPs) having been shown great potential in this field, this paper explores the application of GP regression in forecasting the volatility of foreign exchange returns. This paper builds on the existing literature by applying Gaussian processes for time series forecasting in new ways, namely the multivariate non-coregionalised and coregionalised GP. We show that a multivariate GP can match the accuracy of predictions of a univariate GP, with the added benefit of lower predictive uncertainty due to the incorporation of extra information. Furthermore, we give insight into the relative strengths of the GP methods with recommendations to the practitioner.
- 1 February 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2, 681-686
An Overview of Gaussian process Regression for Volatility Forecasting
Forecasting financial time series using trading data has held the attention of academics and practitioners due to the complexity of the financial system and the profit it can generate for investors. Although investors aim to achieve a consistent attainment of returns, it is important for investors to understand the concept and measurement of volatility. A higher volatility indicates a wider potential range of future returns. But with higher potential returns, comes a higher potential risk. With Gaussian Processes (GPs) having been shown great potential in this field, this paper explores the application of GP regression in forecasting the volatility of foreign exchange returns. This paper builds on the existing literature by applying Gaussian processes for time series forecasting in new ways, namely the multivariate non-coregionalised and coregionalised GP. We show that a multivariate GP can match the accuracy of predictions of a univariate GP, with the added benefit of lower predictive uncertainty due to the incorporation of extra information. Furthermore, we give insight into the relative strengths of the GP methods with recommendations to the practitioner.