Large quantities of asynchronous event sequence data such as crime records, emergence call logs, and financial transactions are becoming increasingly available from various fields. These event sequences often exhibit both long-term and short-term temporal dependencies. Variations of neural network based temporal point processes have been widely used for modeling such asynchronous event sequences. However, many current architectures including attention based point processes struggle with long event sequences due to computational inefficiency. To tackle the challenge, we propose an efficient sparse transformer Hawkes process (STHP), which has two components. For the first component, a transformer with a novel temporal sparse self-attention mechanism is applied to event sequences with arbitrary intervals, mainly focusing on short-term dependencies. For the second component, a transformer is applied to the time series of aggregated event counts, primarily targeting the extraction of long-term periodic dependencies. Both components complement each other and are fused together to model the conditional intensity function of a point process for future event forecasting. Experiments on real-world datasets show that the proposed STHP outperforms baselines and achieves significant improvement in computational efficiency without sacrificing prediction performance for long sequences.
more »
« less
Transformer Hawkes Process
Modern data acquisition routinely produce massive amounts of event sequence data in various domains, such as social media, healthcare, and financial markets. These data often ex- hibit complicated short-term and long-term temporal dependencies. However, most of the ex- isting recurrent neural network-based point process models fail to capture such dependencies, and yield unreliable prediction performance. To address this issue, we propose a Transformer Hawkes Process (THP) model, which leverages the self-attention mechanism to capture long- term dependencies and meanwhile enjoys computational efficiency. Numerical experiments on various datasets show that THP outperforms existing models in terms of both likelihood and event prediction accuracy by a notable margin. Moreover, THP is quite general and can incorpo- rate additional structural knowledge. We provide a concrete example, where THP achieves im- proved prediction performance for learning multiple point processes when incorporating their relational information.
more »
« less
- Award ID(s):
- 1717916
- NSF-PAR ID:
- 10162616
- Date Published:
- Journal Name:
- International Conference on Machine Learning
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Large quantities of asynchronous event sequence data such as crime records, emergence call logs, and financial transactions are becoming increasingly available from various fields. These event sequences often exhibit both long-term and short-term temporal dependencies. Variations of neural network based temporal point processes have been widely used for modeling such asynchronous event sequences. However, many current architectures including attention based point processes struggle with long event sequences due to computational inefficiency. To tackle the challenge, we propose an efficient sparse transformer Hawkes process (STHP), which has two components. For the first component, a transformer with a novel temporal sparse self-attention mechanism is applied to event sequences with arbitrary intervals, mainly focusing on short-term dependencies. For the second component, a transformer is applied to the time series of aggregated event counts, primarily targeting the extraction of long-term periodic dependencies. Both components complement each other and are fused together to model the conditional intensity function of a point process for future event forecasting. Experiments on real-world datasets show that the proposed STHP outperforms baselines and achieves significant improvement in computational efficiency without sacrificing prediction performance for long sequences.more » « less
-
more » « less
Abstract Flood nowcasting refers to near-future prediction of flood status as an extreme weather event unfolds to enhance situational awareness. The objective of this study was to adopt and test a novel structured deep-learning model for urban flood nowcasting by integrating physics-based and human-sensed features. We present a new computational modeling framework including an attention-based spatial–temporal graph convolution network (ASTGCN) model and different streams of data that are collected in real-time, preprocessed, and fed into the model to consider spatial and temporal information and dependencies that improve flood nowcasting. The novelty of the computational modeling framework is threefold: first, the model is capable of considering spatial and temporal dependencies in inundation propagation thanks to the spatial and temporal graph convolutional modules; second, it enables capturing the influence of heterogeneous temporal data streams that can signal flooding status, including physics-based features (e.g., rainfall intensity and water elevation) and human-sensed data (e.g., residents’ flood reports and fluctuations of human activity) on flood nowcasting. Third, its attention mechanism enables the model to direct its focus to the most influential features that vary dynamically and influence the flood nowcasting. We show the application of the modeling framework in the context of Harris County, Texas, as the study area and 2017 Hurricane Harvey as the flood event. Three categories of features are used for nowcasting the extent of flood inundation in different census tracts: (i) static features that capture spatial characteristics of various locations and influence their flood status similarity, (ii) physics-based dynamic features that capture changes in hydrodynamic variables, and (iii) heterogeneous human-sensed dynamic features that capture various aspects of residents’ activities that can provide information regarding flood status. Results indicate that the ASTGCN model provides superior performance for nowcasting of urban flood inundation at the census-tract level, with precision 0.808 and recall 0.891, which shows the model performs better compared with other state-of-the-art models. Moreover, ASTGCN model performance improves when heterogeneous dynamic features are added into the model that solely relies on physics-based features, which demonstrates the promise of using heterogenous human-sensed data for flood nowcasting. Given the results of the comparisons of the models, the proposed modeling framework has the potential to be more investigated when more data of historical events are available in order to develop a predictive tool to provide community responders with an enhanced prediction of the flood inundation during urban flood.
-
Estimating the future event sequence conditioned on current observations is a long-standing and challenging task in temporal analysis. On one hand for many real-world problems the underlying dynamics can be very complex and often unknown. This renders the traditional parametric point process models often fail to fit the data for their limited capacity. On the other hand, long-term prediction suffers from the problem of bias exposure where the error accumulates and propagates to future prediction. Our new model builds upon the sequence to sequence (seq2seq) prediction network. Compared with parametric point process models, its modeling capacity is higher and has better flexibility for fitting real-world data. The main novelty of the paper is to mitigate the second challenge by introducing the likelihood-free loss based on Wasserstein distance between point processes, besides negative maximum likelihood loss used in the traditional seq2seq model. Wasserstein distance, unlike KL divergence i.e. MLE loss, is sensitive to the underlying geometry between samples and can robustly enforce close geometry structure between them. This technique is proven able to improve the vanilla seq2seq model by a notable margin on various tasks.more » « less
-
According to the National Academies, a week long forecast of velocity, vertical structure, and duration of the Loop Current (LC) and its eddies at a given location is a critical step toward understanding their effects on the gulf ecosystems as well as toward anticipating and mitigating the outcomes of anthropogenic and natural disasters in the Gulf of Mexico (GoM). However, creating such a forecast has remained a challenging problem since LC behavior is dominated by dynamic processes across multiple time and spatial scales not resolved at once by conventional numerical models. In this paper, building on the foundation of spatiotemporal predictive learning in video prediction, we develop a physics informed deep learning based prediction model called—Physics-informed Tensor-train ConvLSTM (PITT-ConvLSTM)—for forecasting 3D geo-spatiotemporal sequences. Specifically, we propose (1) a novel 4D higher-order recurrent neural network with empirical orthogonal function analysis to capture the hidden uncorrelated patterns of each hierarchy, (2) a convolutional tensor-train decomposition to capture higher-order space-time correlations, and (3) a mechanism that incorporates prior physics from domain experts by informing the learning in latent space. The advantage of our proposed approach is clear: constrained by the law of physics, the prediction model simultaneously learns good representations for frame dependencies (both short-term and long-term high-level dependency) and inter-hierarchical relations within each time frame. Experiments on geo-spatiotemporal data collected from the GoM demonstrate that the PITT-ConvLSTM model can successfully forecast the volumetric velocity of the LC and its eddies for a period greater than 1 week.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
