This document summarizes a research paper that proposes a deep learning framework called Spatial-Temporal Dynamic Network (STDN) for traffic prediction. STDN uses two key mechanisms: 1) A flow gating mechanism to explicitly model dynamic spatial similarity between locations based on traffic flow data. 2) A periodically shifted attention mechanism to capture long-term periodic dependency while accounting for temporal shifting in traffic patterns. The paper evaluates STDN on real-world taxi and bike-sharing datasets, finding it outperforms other state-of-the-art methods for traffic prediction.
The document analyzes how topography has affected urban sprawl in Beijing over 20 years from 1993 to 2012. It finds that the flat central and eastern parts of Beijing experienced the most development, while the mountainous western and northern areas saw slower growth. Urban areas and nighttime lights expanded fastest towards the flatter land, closely following the development of transportation networks. Topography is thus shown to be a major factor influencing the uneven pattern of Beijing's urbanization.
Deep graph convolutional networks for incident driven traffic speed predictionivaderivader
The document proposes a Deep Incident-Aware Graph Convolutional Network (DIGC-Net) model to predict traffic speeds based on incidents. DIGC-Net uses a graph convolutional network to capture spatial road features, an LSTM to capture temporal patterns, and a fully connected layer to capture periodic features. It also uses an RNN to model how past incidents affect future traffic and speeds. The model discovers critical incidents using anomalous degree and speed variation, extracts latent incident impact features using a classifier, and incorporates these along with spatio-temporal and periodic features for traffic speed prediction.
This document provides an overview of Bayesian inference and filtering techniques for multi-object filtering and multi-target tracking (MOFT). It describes the problems of state estimation in dynamic systems where the state is hidden but can be partially observed. It introduces Bayesian filtering approaches which construct the posterior probability density function of the state using noisy measurements. Specific filtering techniques covered include the Kalman filter, extended Kalman filter, unscented Kalman filter, and particle filter, which make different approximations to solve the filtering problem.
DEEP LEARNING NEURAL NETWORK APPROACHES TO LAND USE-DEMOGRAPHIC- TEMPORAL BA...civejjour
Land use and transportation planning are inter-dependent, as well as being important factors in forecasting urban development. In recent years, predicting traffic based on land use, along with several other variables, has become a worthwhile area of study. In this paper, it is proposed that Deep Neural Network Regression (DNN-Regression) and Recurrent Neural Network (DNN-RNN) methods could be used to predict traffic. These methods used three key variables: land use, demographic and temporal data. The proposed methods were evaluated with other methods, using datasets collected from the City of Calgary, Canada. The proposed DNN-Regression focused on demographic and land use variables for traffic prediction. The study also predicted traffic temporally in the same geographical area by using DNN-RNN. The DNN-RNN used long short-term memory to predict traffic. Comparative experiments revealed that the proposed DNN-Regression and DNN-RNN models outperformed other methods.
Prediction of traveller information and route choiceayishairshad
ayisha irshad ppt Subjected presentation is based on a research paper by
Afzal Ahmeda, Dong Ngoduya & David Watlinga
a Institute for Transport Studies, University of Leeds, 34–40
University Road, Leeds LS2 9JT, UK Published online: 10 Jun 2015.
Deep Multi-View Spatial-Temporal Network for Taxi Demand Predictionivaderivader
This paper proposes DMVST-Net, a deep learning framework that uses three views (spatial, temporal, and semantic) to predict taxi demand. It uses a local CNN to model spatial relationships between nearby regions, an LSTM to model temporal patterns over time, and region embeddings to model semantic relationships between spatially distant but correlated regions. An experiment on a large Didi Chuxing taxi dataset showed DMVST-Net outperformed other methods at predicting future demand, demonstrating the benefit of jointly modeling spatial, temporal, and semantic relationships.
Traffic Prediction from Street Network images.pptxchirantanGupta1
While considering the spatial and temporal features of traffic, capturing the impacts of various external factors on travel is an essential step towards achieving accurate traffic forecasting. However, existing studies seldom consider external factors or neglect the effect of the complex correlations among external factors on traffic. Intuitively, knowledge graphs can naturally describe these correlations. Since knowledge graphs and traffic networks are essentially heterogeneous networks, it is challenging to integrate the information in both networks. On this background, this study presents a knowledge representation-driven traffic forecasting method based on spatial-temporal graph convolutional networks.
The document analyzes how topography has affected urban sprawl in Beijing over 20 years from 1993 to 2012. It finds that the flat central and eastern parts of Beijing experienced the most development, while the mountainous western and northern areas saw slower growth. Urban areas and nighttime lights expanded fastest towards the flatter land, closely following the development of transportation networks. Topography is thus shown to be a major factor influencing the uneven pattern of Beijing's urbanization.
Deep graph convolutional networks for incident driven traffic speed predictionivaderivader
The document proposes a Deep Incident-Aware Graph Convolutional Network (DIGC-Net) model to predict traffic speeds based on incidents. DIGC-Net uses a graph convolutional network to capture spatial road features, an LSTM to capture temporal patterns, and a fully connected layer to capture periodic features. It also uses an RNN to model how past incidents affect future traffic and speeds. The model discovers critical incidents using anomalous degree and speed variation, extracts latent incident impact features using a classifier, and incorporates these along with spatio-temporal and periodic features for traffic speed prediction.
This document provides an overview of Bayesian inference and filtering techniques for multi-object filtering and multi-target tracking (MOFT). It describes the problems of state estimation in dynamic systems where the state is hidden but can be partially observed. It introduces Bayesian filtering approaches which construct the posterior probability density function of the state using noisy measurements. Specific filtering techniques covered include the Kalman filter, extended Kalman filter, unscented Kalman filter, and particle filter, which make different approximations to solve the filtering problem.
DEEP LEARNING NEURAL NETWORK APPROACHES TO LAND USE-DEMOGRAPHIC- TEMPORAL BA...civejjour
Land use and transportation planning are inter-dependent, as well as being important factors in forecasting urban development. In recent years, predicting traffic based on land use, along with several other variables, has become a worthwhile area of study. In this paper, it is proposed that Deep Neural Network Regression (DNN-Regression) and Recurrent Neural Network (DNN-RNN) methods could be used to predict traffic. These methods used three key variables: land use, demographic and temporal data. The proposed methods were evaluated with other methods, using datasets collected from the City of Calgary, Canada. The proposed DNN-Regression focused on demographic and land use variables for traffic prediction. The study also predicted traffic temporally in the same geographical area by using DNN-RNN. The DNN-RNN used long short-term memory to predict traffic. Comparative experiments revealed that the proposed DNN-Regression and DNN-RNN models outperformed other methods.
Prediction of traveller information and route choiceayishairshad
ayisha irshad ppt Subjected presentation is based on a research paper by
Afzal Ahmeda, Dong Ngoduya & David Watlinga
a Institute for Transport Studies, University of Leeds, 34–40
University Road, Leeds LS2 9JT, UK Published online: 10 Jun 2015.
Deep Multi-View Spatial-Temporal Network for Taxi Demand Predictionivaderivader
This paper proposes DMVST-Net, a deep learning framework that uses three views (spatial, temporal, and semantic) to predict taxi demand. It uses a local CNN to model spatial relationships between nearby regions, an LSTM to model temporal patterns over time, and region embeddings to model semantic relationships between spatially distant but correlated regions. An experiment on a large Didi Chuxing taxi dataset showed DMVST-Net outperformed other methods at predicting future demand, demonstrating the benefit of jointly modeling spatial, temporal, and semantic relationships.
Traffic Prediction from Street Network images.pptxchirantanGupta1
While considering the spatial and temporal features of traffic, capturing the impacts of various external factors on travel is an essential step towards achieving accurate traffic forecasting. However, existing studies seldom consider external factors or neglect the effect of the complex correlations among external factors on traffic. Intuitively, knowledge graphs can naturally describe these correlations. Since knowledge graphs and traffic networks are essentially heterogeneous networks, it is challenging to integrate the information in both networks. On this background, this study presents a knowledge representation-driven traffic forecasting method based on spatial-temporal graph convolutional networks.
Coupled Layer-wise Graph Convolution for Transportation Demand Predictionivaderivader
This document summarizes a research paper that proposes a Coupled Layer-wise Convolutional Recurrent Neural Network (CCRNN) model for transportation demand prediction. CCRNN uses the following key techniques:
1. It generates adjacency matrices at each layer of the model to capture multi-level spatial dependencies, rather than using a fixed adjacency matrix.
2. It employs coupled layer-wise graph convolutions with different adjacency matrices at each layer to obtain representations at different levels of abstraction.
3. A multi-level attention mechanism aggregates representations from different layers.
4. A gated recurrent unit with graph convolutions models temporal dependencies over time.
The paper experiments with CCRNN on two
An effective joint prediction model for travel demands and traffic flowsivaderivader
This document summarizes a research paper that presents DeepTP, a joint prediction model for travel demands and traffic flows. DeepTP uses four modules: 1) a future spatio-temporal encoding module, 2) a past traffic sequence encoding module, 3) a graph-based correlation encoding module, and 4) a final estimation module. It encodes three types of embeddings - past traffic data, region-level correlations, and temporal periodicity - to capture inter-traffic correlations, region-level similarities, and periodic patterns in demand and flow. The model was evaluated on real-world traffic datasets from two cities and was shown to outperform other baselines in joint demand and flow prediction.
The document proposes a novel deep learning framework called Spatio-Temporal Graph Convolutional Networks (STGCN) to tackle the time series prediction problem in traffic forecasting. STGCN uses graph convolutional layers to model spatial dependencies on a traffic network represented as a graph, and convolutional layers to model temporal dependencies. Experiments show STGCN outperforms state-of-the-art baselines by effectively capturing comprehensive spatio-temporal correlations through modeling multi-scale traffic networks.
[20240318_LabSeminar_Huy]GSTNet: Global Spatial-Temporal Network for Traffic ...thanhdowork
GSTNet is a deep learning model for traffic flow prediction that incorporates spatial and temporal information. It contains multi-resolution temporal and global correlated spatial modules. The temporal module captures short and long-term patterns, while the spatial module considers both local and non-local correlations between locations. In experiments on Beijing transportation data, GSTNet achieved more accurate predictions compared to other methods and was able to capture both short and long-term dependencies in traffic flow.
This document describes a new spatial-temporal attention graph convolution network (STAGCN) for traffic forecasting. STAGCN contains three key components: 1) a graph learning layer that learns both a static graph capturing global spatial relationships and a dynamic graph capturing local spatial changes, 2) an adaptive graph convolution layer, and 3) a gated temporal attention module using causal-trend attention to model long-term temporal dependencies by focusing on important historical information. Experimental results on four traffic datasets show STAGCN achieves better prediction accuracy than existing methods.
This document summarizes and compares several deep learning models for traffic speed prediction:
- DCRNN was the first model to use GCN in the traffic domain, combining GCN with RNN. STGCN used a many-to-one architecture with GCN and CNN 12 times to predict 12 time sequences.
- ASTGCN used attention mechanisms with GCN and CNN, applying spatial and temporal attention. STSGCN simultaneously captured spatial and temporal features using a localized spatial-temporal graph with GCN.
- Graph-WaveNet used GCN and dilated CNN with an adaptive adjacency matrix. Additional experiments are needed to fairly evaluate models and ablate contributions of components.
The document proposes a method for traffic flow prediction using KNN and LSTM. KNN is used to select neighboring stations that are spatially correlated with the test station. A two-layer LSTM network then predicts traffic flow in the selected stations. The predictions are combined using weighted averaging to obtain the final prediction for the test station. An experiment on traffic data from Minnesota highways found the proposed method improved prediction accuracy over ARIMA, SVR, WNN, DBN-SVR, and LSTM-only models, achieving on average a 12.59% reduction in error.
The document proposes a new framework called Spatial-Temporal Synchronous Graph Convolutional Networks (STSGCN) for spatial-temporal network data forecasting. STSGCN is able to capture complex localized spatial-temporal correlations through an elaborately designed spatial-temporal synchronous modeling mechanism. It constructs localized spatial-temporal graphs by connecting individual spatial graphs of adjacent time steps. STSGCN then uses a Spatial-Temporal Synchronous Graph Convolutional Module to directly capture the localized spatial-temporal correlations in these graphs. It also includes multiple modules to effectively model heterogeneities in different time periods. Extensive experiments on four real-world datasets demonstrate that STSGCN achieves state-of-the-art performance.
Traffic Flow Forecasting with Spatial-Temporal Graph Diffusion Networkivaderivader
The document summarizes the key aspects of the paper "Traffic Flow Forecasting with Spatial-Temporal Graph Diffusion Network". It introduces the ST-GDN framework which uses a hierarchical graph neural architecture to model temporal hierarchies and learn traffic dependencies within and across regions via graph diffusion. The methodology section explains the multi-scale temporal modeling, global context learning using attention mechanisms, and region-wise relation encoding on the graph. The experiment section compares ST-GDN to other baselines on taxi and bike traffic datasets, finding ST-GDN achieves better performance. The conclusion discusses potential extensions like using different clustering and adjacency matrix approaches.
Deep Learning Neural Network Approaches to Land Use-demographic- Temporal bas...civejjour
This document describes research that uses deep learning neural network approaches to predict traffic based on land use, demographic, and temporal data. Specifically, it proposes using Deep Neural Network Regression (DNN-Regression) and Recurrent Neural Network (DNN-RNN) models. DNN-Regression focuses on demographic and land use variables to predict traffic, while DNN-RNN predicts traffic temporally in the same geographic area using long short-term memory. The models are evaluated using datasets from Calgary, Canada, and are found to outperform other existing approaches to traffic prediction.
[20240325_LabSeminar_Huy]Spatial-Temporal Fusion Graph Neural Networks for Tr...thanhdowork
This document outlines a spatial-temporal graph neural network model called STFGN for traffic flow forecasting. STFGN constructs a temporal graph using dynamic time warping to capture temporal dependencies beyond spatial proximity. It then builds a fusion graph combining the spatial, temporal, and temporal connectivity graphs. The model uses STFGN modules and gated dilated CNNs to learn local and global spatial-temporal patterns. Experiments on California traffic datasets show STFGN achieves state-of-the-art performance for traffic flow prediction.
The document proposes an Attention Temporal Graph Convolutional Network (A3T-GCN) model for traffic forecasting that aims to simultaneously capture global temporal dynamics and spatial correlations in traffic data. The A3T-GCN combines a Graph Convolutional Network (GCN) to learn spatial dependencies based on road network topology with a Gated Recurrent Unit (GRU) to learn temporal trends, and introduces an attention mechanism to adjust the importance of different time points and better predict traffic. Experimental results on real-world datasets demonstrate the effectiveness of the proposed A3T-GCN model.
This document summarizes a survey of traffic data visualization techniques. It describes how traffic data is collected from sensors and organized. Common visualization methods are discussed like using lines and regions to show trajectories and aggregated spatial data, and space-time cubes to visualize trajectories over time. The document also outlines techniques for visualizing multiple attributes, analyzing patterns and clustering trajectories, monitoring traffic situations, and exploring data. Future work opportunities are developing systems for big data, situation awareness, and integrating different visualization views.
In this short presentation, we will provide some recent developments in the field of crowd monitoring, modelling and management. We will illustrate these by showing various projects that we are involved in, including the SmartStation project, and the different events organised in and around the city of Amsterdam (including the Europride, SAIL, etc.).
In the talk, we will discuss the different components of the system and the methods and technology involved in these. We focus on advanced data collection techniques, the use of social media data, data fusion and the advanced macroscopic modelling required for this. Also, we will show examples of interventions that have been tested, showing how these systems are used in practise.
This document proposes a new graph neural network model called Spatio-Temporal Pivotal Graph Neural Networks (STPGNN) for traffic flow forecasting. STPGNN first identifies pivotal nodes in the traffic network that exhibit extensive connections using a scoring mechanism. It then constructs a pivotal graph and uses a novel pivotal graph convolution module to capture spatio-temporal dependencies centered around these pivotal nodes. Experiments on several traffic datasets show STPGNN achieves better forecasting accuracy compared to other methods while maintaining lower computational cost.
The document discusses approaches for providing accurate and robust traffic forecasts using empirical data. It describes two main types of empirical approaches: basic forecast approaches that use individual prediction models, and combined forecast approaches that combine different forecasts into a single prediction. Basic approaches can be parametric (e.g. linear regression) or non-parametric (e.g. neural networks), while combined approaches aim to improve accuracy by incorporating each model's unique strengths. The document provides an overview of various techniques within each category.
Prediction and planning for self driving at waymoYu Huang
ChauffeurNet: Learning To Drive By Imitating The Best Synthesizing The Worst
Multipath: Multiple Probabilistic Anchor Trajectory Hypotheses For Behavior Prediction
VectorNet: Encoding HD Maps And Agent Dynamics From Vectorized Representation
TNT: Target-driven Trajectory Prediction
Large Scale Interactive Motion Forecasting For Autonomous Driving : The Waymo Open Motion Dataset
Identifying Driver Interactions Via Conditional Behavior Prediction
Peeking Into The Future: Predicting Future Person Activities And Locations In Videos
STINet: Spatio-temporal-interactive Network For Pedestrian Detection And Trajectory Prediction
This document summarizes research presented at recent Argument Mining workshops. It describes tasks from 2021 and 2022 on key point analysis and validity-novelty prediction. It also outlines papers on extracting argument components, predicting reasoning markers, and detecting argument boundaries. The document indicates that argument mining can analyze debates on social media and that legal corpora can help develop argument mining systems.
This document summarizes several papers being presented at the CHI 2023 conference related to human-AI collaboration, mental healthcare, comment visualization, and virtual reality. For human-AI collaboration, one paper studies the effects of AI code generators on novice programmers and another aims to bridge the abstraction gap between end users and large language models. A third paper investigates whether groups or individuals are better at AI-assisted decision making. For mental healthcare, two papers study user perceptions of recommender systems and interest in self-tracking technologies. For comment visualization, a paper examines selective avoidance of uncongenial facts. Three papers on virtual reality discuss opportunities for metaverse workspaces, embodying physics-aware avatars, and mapping finger motions
Coupled Layer-wise Graph Convolution for Transportation Demand Predictionivaderivader
This document summarizes a research paper that proposes a Coupled Layer-wise Convolutional Recurrent Neural Network (CCRNN) model for transportation demand prediction. CCRNN uses the following key techniques:
1. It generates adjacency matrices at each layer of the model to capture multi-level spatial dependencies, rather than using a fixed adjacency matrix.
2. It employs coupled layer-wise graph convolutions with different adjacency matrices at each layer to obtain representations at different levels of abstraction.
3. A multi-level attention mechanism aggregates representations from different layers.
4. A gated recurrent unit with graph convolutions models temporal dependencies over time.
The paper experiments with CCRNN on two
An effective joint prediction model for travel demands and traffic flowsivaderivader
This document summarizes a research paper that presents DeepTP, a joint prediction model for travel demands and traffic flows. DeepTP uses four modules: 1) a future spatio-temporal encoding module, 2) a past traffic sequence encoding module, 3) a graph-based correlation encoding module, and 4) a final estimation module. It encodes three types of embeddings - past traffic data, region-level correlations, and temporal periodicity - to capture inter-traffic correlations, region-level similarities, and periodic patterns in demand and flow. The model was evaluated on real-world traffic datasets from two cities and was shown to outperform other baselines in joint demand and flow prediction.
The document proposes a novel deep learning framework called Spatio-Temporal Graph Convolutional Networks (STGCN) to tackle the time series prediction problem in traffic forecasting. STGCN uses graph convolutional layers to model spatial dependencies on a traffic network represented as a graph, and convolutional layers to model temporal dependencies. Experiments show STGCN outperforms state-of-the-art baselines by effectively capturing comprehensive spatio-temporal correlations through modeling multi-scale traffic networks.
[20240318_LabSeminar_Huy]GSTNet: Global Spatial-Temporal Network for Traffic ...thanhdowork
GSTNet is a deep learning model for traffic flow prediction that incorporates spatial and temporal information. It contains multi-resolution temporal and global correlated spatial modules. The temporal module captures short and long-term patterns, while the spatial module considers both local and non-local correlations between locations. In experiments on Beijing transportation data, GSTNet achieved more accurate predictions compared to other methods and was able to capture both short and long-term dependencies in traffic flow.
This document describes a new spatial-temporal attention graph convolution network (STAGCN) for traffic forecasting. STAGCN contains three key components: 1) a graph learning layer that learns both a static graph capturing global spatial relationships and a dynamic graph capturing local spatial changes, 2) an adaptive graph convolution layer, and 3) a gated temporal attention module using causal-trend attention to model long-term temporal dependencies by focusing on important historical information. Experimental results on four traffic datasets show STAGCN achieves better prediction accuracy than existing methods.
This document summarizes and compares several deep learning models for traffic speed prediction:
- DCRNN was the first model to use GCN in the traffic domain, combining GCN with RNN. STGCN used a many-to-one architecture with GCN and CNN 12 times to predict 12 time sequences.
- ASTGCN used attention mechanisms with GCN and CNN, applying spatial and temporal attention. STSGCN simultaneously captured spatial and temporal features using a localized spatial-temporal graph with GCN.
- Graph-WaveNet used GCN and dilated CNN with an adaptive adjacency matrix. Additional experiments are needed to fairly evaluate models and ablate contributions of components.
The document proposes a method for traffic flow prediction using KNN and LSTM. KNN is used to select neighboring stations that are spatially correlated with the test station. A two-layer LSTM network then predicts traffic flow in the selected stations. The predictions are combined using weighted averaging to obtain the final prediction for the test station. An experiment on traffic data from Minnesota highways found the proposed method improved prediction accuracy over ARIMA, SVR, WNN, DBN-SVR, and LSTM-only models, achieving on average a 12.59% reduction in error.
The document proposes a new framework called Spatial-Temporal Synchronous Graph Convolutional Networks (STSGCN) for spatial-temporal network data forecasting. STSGCN is able to capture complex localized spatial-temporal correlations through an elaborately designed spatial-temporal synchronous modeling mechanism. It constructs localized spatial-temporal graphs by connecting individual spatial graphs of adjacent time steps. STSGCN then uses a Spatial-Temporal Synchronous Graph Convolutional Module to directly capture the localized spatial-temporal correlations in these graphs. It also includes multiple modules to effectively model heterogeneities in different time periods. Extensive experiments on four real-world datasets demonstrate that STSGCN achieves state-of-the-art performance.
Traffic Flow Forecasting with Spatial-Temporal Graph Diffusion Networkivaderivader
The document summarizes the key aspects of the paper "Traffic Flow Forecasting with Spatial-Temporal Graph Diffusion Network". It introduces the ST-GDN framework which uses a hierarchical graph neural architecture to model temporal hierarchies and learn traffic dependencies within and across regions via graph diffusion. The methodology section explains the multi-scale temporal modeling, global context learning using attention mechanisms, and region-wise relation encoding on the graph. The experiment section compares ST-GDN to other baselines on taxi and bike traffic datasets, finding ST-GDN achieves better performance. The conclusion discusses potential extensions like using different clustering and adjacency matrix approaches.
Deep Learning Neural Network Approaches to Land Use-demographic- Temporal bas...civejjour
This document describes research that uses deep learning neural network approaches to predict traffic based on land use, demographic, and temporal data. Specifically, it proposes using Deep Neural Network Regression (DNN-Regression) and Recurrent Neural Network (DNN-RNN) models. DNN-Regression focuses on demographic and land use variables to predict traffic, while DNN-RNN predicts traffic temporally in the same geographic area using long short-term memory. The models are evaluated using datasets from Calgary, Canada, and are found to outperform other existing approaches to traffic prediction.
[20240325_LabSeminar_Huy]Spatial-Temporal Fusion Graph Neural Networks for Tr...thanhdowork
This document outlines a spatial-temporal graph neural network model called STFGN for traffic flow forecasting. STFGN constructs a temporal graph using dynamic time warping to capture temporal dependencies beyond spatial proximity. It then builds a fusion graph combining the spatial, temporal, and temporal connectivity graphs. The model uses STFGN modules and gated dilated CNNs to learn local and global spatial-temporal patterns. Experiments on California traffic datasets show STFGN achieves state-of-the-art performance for traffic flow prediction.
The document proposes an Attention Temporal Graph Convolutional Network (A3T-GCN) model for traffic forecasting that aims to simultaneously capture global temporal dynamics and spatial correlations in traffic data. The A3T-GCN combines a Graph Convolutional Network (GCN) to learn spatial dependencies based on road network topology with a Gated Recurrent Unit (GRU) to learn temporal trends, and introduces an attention mechanism to adjust the importance of different time points and better predict traffic. Experimental results on real-world datasets demonstrate the effectiveness of the proposed A3T-GCN model.
This document summarizes a survey of traffic data visualization techniques. It describes how traffic data is collected from sensors and organized. Common visualization methods are discussed like using lines and regions to show trajectories and aggregated spatial data, and space-time cubes to visualize trajectories over time. The document also outlines techniques for visualizing multiple attributes, analyzing patterns and clustering trajectories, monitoring traffic situations, and exploring data. Future work opportunities are developing systems for big data, situation awareness, and integrating different visualization views.
In this short presentation, we will provide some recent developments in the field of crowd monitoring, modelling and management. We will illustrate these by showing various projects that we are involved in, including the SmartStation project, and the different events organised in and around the city of Amsterdam (including the Europride, SAIL, etc.).
In the talk, we will discuss the different components of the system and the methods and technology involved in these. We focus on advanced data collection techniques, the use of social media data, data fusion and the advanced macroscopic modelling required for this. Also, we will show examples of interventions that have been tested, showing how these systems are used in practise.
This document proposes a new graph neural network model called Spatio-Temporal Pivotal Graph Neural Networks (STPGNN) for traffic flow forecasting. STPGNN first identifies pivotal nodes in the traffic network that exhibit extensive connections using a scoring mechanism. It then constructs a pivotal graph and uses a novel pivotal graph convolution module to capture spatio-temporal dependencies centered around these pivotal nodes. Experiments on several traffic datasets show STPGNN achieves better forecasting accuracy compared to other methods while maintaining lower computational cost.
The document discusses approaches for providing accurate and robust traffic forecasts using empirical data. It describes two main types of empirical approaches: basic forecast approaches that use individual prediction models, and combined forecast approaches that combine different forecasts into a single prediction. Basic approaches can be parametric (e.g. linear regression) or non-parametric (e.g. neural networks), while combined approaches aim to improve accuracy by incorporating each model's unique strengths. The document provides an overview of various techniques within each category.
Prediction and planning for self driving at waymoYu Huang
ChauffeurNet: Learning To Drive By Imitating The Best Synthesizing The Worst
Multipath: Multiple Probabilistic Anchor Trajectory Hypotheses For Behavior Prediction
VectorNet: Encoding HD Maps And Agent Dynamics From Vectorized Representation
TNT: Target-driven Trajectory Prediction
Large Scale Interactive Motion Forecasting For Autonomous Driving : The Waymo Open Motion Dataset
Identifying Driver Interactions Via Conditional Behavior Prediction
Peeking Into The Future: Predicting Future Person Activities And Locations In Videos
STINet: Spatio-temporal-interactive Network For Pedestrian Detection And Trajectory Prediction
This document summarizes research presented at recent Argument Mining workshops. It describes tasks from 2021 and 2022 on key point analysis and validity-novelty prediction. It also outlines papers on extracting argument components, predicting reasoning markers, and detecting argument boundaries. The document indicates that argument mining can analyze debates on social media and that legal corpora can help develop argument mining systems.
This document summarizes several papers being presented at the CHI 2023 conference related to human-AI collaboration, mental healthcare, comment visualization, and virtual reality. For human-AI collaboration, one paper studies the effects of AI code generators on novice programmers and another aims to bridge the abstraction gap between end users and large language models. A third paper investigates whether groups or individuals are better at AI-assisted decision making. For mental healthcare, two papers study user perceptions of recommender systems and interest in self-tracking technologies. For comment visualization, a paper examines selective avoidance of uncongenial facts. Three papers on virtual reality discuss opportunities for metaverse workspaces, embodying physics-aware avatars, and mapping finger motions
DDGK: Learning Graph Representations for Deep Divergence Graph Kernelsivaderivader
This document summarizes a research paper on learning graph representations for deep divergence graph kernels (DDGK). DDGK learns graph representations without supervision or domain knowledge by using a node-to-edges encoder and isomorphism attention. The isomorphism attention provides a bidirectional mapping between nodes in two graphs. DDGK then calculates a divergence score between the source and target graphs as a measure of their (dis)similarity. Experimental results showed DDGK produces representations competitive with other graph kernel baselines. The paper proposes several extensions, including different graph encoders and attention mechanisms, as well as improved regularization and scalability.
So Predictable! Continuous 3D Hand Trajectory Prediction in Virtual Reality ivaderivader
This document summarizes research on developing a hybrid kinematic regression model for continuous 3D hand trajectory prediction in virtual reality. The model combines classical kinematics with regression to predict hand movements up to 300ms into the future with low error. A user study was conducted to collect training data from structured pointing tasks and unstructured VR gameplay. The model was evaluated through cross-validation and testing on new users and activities, demonstrating good performance without retraining. Limitations and opportunities for future work are discussed, such as expanding the model to support two-handed interactions and other types of tasks.
Reinforcement Learning-based Placement of Charging Stations in Urban Road Net...ivaderivader
This document summarizes a research paper that proposes using reinforcement learning to optimize the placement of electric vehicle charging stations in urban road networks. It formulates the charging station placement task as a reinforcement learning problem to maximize a defined utility function. The paper details the implementation of a deep Q-learning agent that takes actions to create, increase, or relocate charging stations. It evaluates the reinforcement learning approach on real-world road networks and charging demand datasets from two German cities, finding it outperforms baseline greedy algorithms in maximizing utility while meeting constraints.
Prediction for Retrospection: Integrating Algorithmic Stress Prediction into ...ivaderivader
This document describes MindScope, an algorithm-assisted stress management system for college students. It collects behavioral data from students' smartphones to predict their stress levels and provides explanations of the predictions. A study was conducted with 36 students who used MindScope for 25 days. It found that providing algorithmic explanations for stress predictions helped reduce students' perceived stress levels over time. It also supported their stress management practices and self-reflection. However, the study was limited by its short duration and lack of a control group.
A Style-Based Generator Architecture for Generative Adversarial Networksivaderivader
StyleGAN is a generative adversarial network that achieves disentangled and scalable image generation. It uses adaptive instance normalization (AdaIN) to modify feature statistics at different scales, allowing scale-specific image stylization. The generator is designed as a learned mapping from latent space to image space. Latent codes are fed into each layer and transformed through AdaIN to modify feature statistics. This disentangles high-level attributes like pose, hair, etc. and allows controllable image synthesis through interpolation in latent space.
Neural Approximate Dynamic Programming for On-Demand Ride-Poolingivaderivader
This document summarizes a research paper on using neural approximate dynamic programming (NeurADP) to solve the ride-pooling matching problem (RMP) of optimally assigning passenger requests to vehicles. The key contributions are developing NeurADP, which uses a neural network to approximate the value function within an ADP framework, and connecting it to reinforcement learning. This allows NeurADP to handle the complex RMP involving partially filled vehicles better than prior methods. The researchers test NeurADP on New York City taxi data and find that changing constraints like maximum wait time and vehicle capacity affects NeurADP's ability to serve more requests while avoiding myopic assignments.
Bad Breakdowns, Useful Seams, and Face Slapping: Analysis of VR Fails on YouTubeivaderivader
This document analyzes 233 VR fail videos from YouTube to understand types and causes of VR fails outside the lab. The analysis found that the most common types were excessive reaction (53%) and falling over (18%), while the top causes were fear (40%) and sensorimotor mismatch (26%). Spectators were often laughing or expressing concern. The analysis suggests design implications like changing play spaces, mechanics, and interactions to prevent fails while promoting seamful design and spectator engagement. Limitations include the corpus not being broadly representative and difficulty analyzing home videos.
Invertible Denoising Network: A Light Solution for Real Noise Removalivaderivader
This document summarizes an academic paper on invertible denoising networks (InvDN), a new type of neural network designed for real image denoising. The key points are:
1) InvDN is the first paper to design invertible networks for real image denoising by utilizing two different distributions - one for the noisy image and one for the clean image.
2) InvDN shows state-of-the-art results on benchmark datasets by restoring images and generating new noisy images from the restored clean images.
3) InvDN works by splitting images into low and high frequency components in the frequency domain and handling each separately, allowing the disentanglement of clean image and noise information needed for
Traffic Demand Prediction Based Dynamic Transition Convolutional Neural Networkivaderivader
The document proposes a traffic demand prediction model based on a dynamic transition convolutional neural network. It defines a transition network using density-peak based clustering to identify virtual stations. A dynamic graph convolutional gated recurrent unit is developed to model the station-to-station transitions over time. Meteorological data is also incorporated as external features to improve the prediction performance. The model is evaluated on bike and taxi demand data from New York City and shows improved results over various baseline models.
MusicBERT: Symbolic Music Understanding with Large-Scale Pre-Training ivaderivader
The document introduces MusicBERT, a large-scale pre-trained Transformer model for symbolic music understanding. MusicBERT uses a novel encoding method called OctupleMIDI to efficiently encode music sequences into compact tokens. It also employs a bar-level masking strategy during pre-training. MusicBERT is evaluated on four music understanding tasks and achieves state-of-the-art results, demonstrating the effectiveness of its encoding method and pre-training approach.
Screen2Vec: Semantic Embedding of GUI Screens and GUI Componentsivaderivader
This document presents Screen2Vec, a technique for generating semantic embeddings of GUI screens and components using text, visual design, layout patterns, and app metadata. It has a two-level architecture that learns embeddings at the component and screen levels. The technique is self-supervised and does not require labeled data. It was trained on the RICO dataset and can be used for tasks like nearest neighbor retrieval and representing mobile tasks. Sample results showed Screen2Vec generated more comprehensive representations compared to baselines using only text or layout.
Augmenting Decisions of Taxi Drivers through Reinforcement Learning for Impro...ivaderivader
The document summarizes research on using reinforcement learning to help taxi drivers maximize revenue by recommending optimal locations to drive to at different times and days. The methodology involves modeling driver decisions as activity graphs from taxi trajectory data and then using Monte Carlo reinforcement learning with states defined by zone, time, and day to learn an optimal policy. Experiments show the learned policy is able to achieve higher average revenues than top-earning drivers and common heuristics, demonstrating the potential of reinforcement learning to help taxi drivers.
Natural Language to Visualization by Neural Machine Translationivaderivader
This document summarizes a method called ncNet that uses neural machine translation to translate natural language queries to Vega-Zero visualization specifications. Key points include:
- ncNet is a Transformer-based seq2seq model that can translate natural language queries about data into specifications for visualizations.
- It uses techniques like attention forcing and visualization-aware translation to generate valid Vega-Zero outputs from the queries.
- An evaluation compares ncNet to existing methods on a dataset of natural language queries paired with Vega-Zero specifications, finding it outperforms baselines.
CAKE: Sharing Slices of Confidential Data on BlockchainClaudio Di Ciccio
Presented at the CAiSE 2024 Forum, Intelligent Information Systems, June 6th, Limassol, Cyprus.
Synopsis: Cooperative information systems typically involve various entities in a collaborative process within a distributed environment. Blockchain technology offers a mechanism for automating such processes, even when only partial trust exists among participants. The data stored on the blockchain is replicated across all nodes in the network, ensuring accessibility to all participants. While this aspect facilitates traceability, integrity, and persistence, it poses challenges for adopting public blockchains in enterprise settings due to confidentiality issues. In this paper, we present a software tool named Control Access via Key Encryption (CAKE), designed to ensure data confidentiality in scenarios involving public blockchains. After outlining its core components and functionalities, we showcase the application of CAKE in the context of a real-world cyber-security project within the logistics domain.
Paper: https://doi.org/10.1007/978-3-031-61000-4_16
UiPath Test Automation using UiPath Test Suite series, part 6DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 6. In this session, we will cover Test Automation with generative AI and Open AI.
UiPath Test Automation with generative AI and Open AI webinar offers an in-depth exploration of leveraging cutting-edge technologies for test automation within the UiPath platform. Attendees will delve into the integration of generative AI, a test automation solution, with Open AI advanced natural language processing capabilities.
Throughout the session, participants will discover how this synergy empowers testers to automate repetitive tasks, enhance testing accuracy, and expedite the software testing life cycle. Topics covered include the seamless integration process, practical use cases, and the benefits of harnessing AI-driven automation for UiPath testing initiatives. By attending this webinar, testers, and automation professionals can gain valuable insights into harnessing the power of AI to optimize their test automation workflows within the UiPath ecosystem, ultimately driving efficiency and quality in software development processes.
What will you get from this session?
1. Insights into integrating generative AI.
2. Understanding how this integration enhances test automation within the UiPath platform
3. Practical demonstrations
4. Exploration of real-world use cases illustrating the benefits of AI-driven test automation for UiPath
Topics covered:
What is generative AI
Test Automation with generative AI and Open AI.
UiPath integration with generative AI
Speaker:
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
How to Get CNIC Information System with Paksim Ga.pptxdanishmna97
Pakdata Cf is a groundbreaking system designed to streamline and facilitate access to CNIC information. This innovative platform leverages advanced technology to provide users with efficient and secure access to their CNIC details.
GraphRAG for Life Science to increase LLM accuracyTomaz Bratanic
GraphRAG for life science domain, where you retriever information from biomedical knowledge graphs using LLMs to increase the accuracy and performance of generated answers
Threats to mobile devices are more prevalent and increasing in scope and complexity. Users of mobile devices desire to take full advantage of the features
available on those devices, but many of the features provide convenience and capability but sacrifice security. This best practices guide outlines steps the users can take to better protect personal devices and information.
TrustArc Webinar - 2024 Global Privacy SurveyTrustArc
How does your privacy program stack up against your peers? What challenges are privacy teams tackling and prioritizing in 2024?
In the fifth annual Global Privacy Benchmarks Survey, we asked over 1,800 global privacy professionals and business executives to share their perspectives on the current state of privacy inside and outside of their organizations. This year’s report focused on emerging areas of importance for privacy and compliance professionals, including considerations and implications of Artificial Intelligence (AI) technologies, building brand trust, and different approaches for achieving higher privacy competence scores.
See how organizational priorities and strategic approaches to data security and privacy are evolving around the globe.
This webinar will review:
- The top 10 privacy insights from the fifth annual Global Privacy Benchmarks Survey
- The top challenges for privacy leaders, practitioners, and organizations in 2024
- Key themes to consider in developing and maintaining your privacy program
Monitoring and Managing Anomaly Detection on OpenShift.pdfTosin Akinosho
Monitoring and Managing Anomaly Detection on OpenShift
Overview
Dive into the world of anomaly detection on edge devices with our comprehensive hands-on tutorial. This SlideShare presentation will guide you through the entire process, from data collection and model training to edge deployment and real-time monitoring. Perfect for those looking to implement robust anomaly detection systems on resource-constrained IoT/edge devices.
Key Topics Covered
1. Introduction to Anomaly Detection
- Understand the fundamentals of anomaly detection and its importance in identifying unusual behavior or failures in systems.
2. Understanding Edge (IoT)
- Learn about edge computing and IoT, and how they enable real-time data processing and decision-making at the source.
3. What is ArgoCD?
- Discover ArgoCD, a declarative, GitOps continuous delivery tool for Kubernetes, and its role in deploying applications on edge devices.
4. Deployment Using ArgoCD for Edge Devices
- Step-by-step guide on deploying anomaly detection models on edge devices using ArgoCD.
5. Introduction to Apache Kafka and S3
- Explore Apache Kafka for real-time data streaming and Amazon S3 for scalable storage solutions.
6. Viewing Kafka Messages in the Data Lake
- Learn how to view and analyze Kafka messages stored in a data lake for better insights.
7. What is Prometheus?
- Get to know Prometheus, an open-source monitoring and alerting toolkit, and its application in monitoring edge devices.
8. Monitoring Application Metrics with Prometheus
- Detailed instructions on setting up Prometheus to monitor the performance and health of your anomaly detection system.
9. What is Camel K?
- Introduction to Camel K, a lightweight integration framework built on Apache Camel, designed for Kubernetes.
10. Configuring Camel K Integrations for Data Pipelines
- Learn how to configure Camel K for seamless data pipeline integrations in your anomaly detection workflow.
11. What is a Jupyter Notebook?
- Overview of Jupyter Notebooks, an open-source web application for creating and sharing documents with live code, equations, visualizations, and narrative text.
12. Jupyter Notebooks with Code Examples
- Hands-on examples and code snippets in Jupyter Notebooks to help you implement and test anomaly detection models.
Full-RAG: A modern architecture for hyper-personalizationZilliz
Mike Del Balso, CEO & Co-Founder at Tecton, presents "Full RAG," a novel approach to AI recommendation systems, aiming to push beyond the limitations of traditional models through a deep integration of contextual insights and real-time data, leveraging the Retrieval-Augmented Generation architecture. This talk will outline Full RAG's potential to significantly enhance personalization, address engineering challenges such as data management and model training, and introduce data enrichment with reranking as a key solution. Attendees will gain crucial insights into the importance of hyperpersonalization in AI, the capabilities of Full RAG for advanced personalization, and strategies for managing complex data integrations for deploying cutting-edge AI solutions.
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2024/06/building-and-scaling-ai-applications-with-the-nx-ai-manager-a-presentation-from-network-optix/
Robin van Emden, Senior Director of Data Science at Network Optix, presents the “Building and Scaling AI Applications with the Nx AI Manager,” tutorial at the May 2024 Embedded Vision Summit.
In this presentation, van Emden covers the basics of scaling edge AI solutions using the Nx tool kit. He emphasizes the process of developing AI models and deploying them globally. He also showcases the conversion of AI models and the creation of effective edge AI pipelines, with a focus on pre-processing, model conversion, selecting the appropriate inference engine for the target hardware and post-processing.
van Emden shows how Nx can simplify the developer’s life and facilitate a rapid transition from concept to production-ready applications.He provides valuable insights into developing scalable and efficient edge AI solutions, with a strong focus on practical implementation.
AI 101: An Introduction to the Basics and Impact of Artificial IntelligenceIndexBug
Imagine a world where machines not only perform tasks but also learn, adapt, and make decisions. This is the promise of Artificial Intelligence (AI), a technology that's not just enhancing our lives but revolutionizing entire industries.
Generating privacy-protected synthetic data using Secludy and MilvusZilliz
During this demo, the founders of Secludy will demonstrate how their system utilizes Milvus to store and manipulate embeddings for generating privacy-protected synthetic data. Their approach not only maintains the confidentiality of the original data but also enhances the utility and scalability of LLMs under privacy constraints. Attendees, including machine learning engineers, data scientists, and data managers, will witness first-hand how Secludy's integration with Milvus empowers organizations to harness the power of LLMs securely and efficiently.
Ocean lotus Threat actors project by John Sitima 2024 (1).pptxSitimaJohn
Ocean Lotus cyber threat actors represent a sophisticated, persistent, and politically motivated group that poses a significant risk to organizations and individuals in the Southeast Asian region. Their continuous evolution and adaptability underscore the need for robust cybersecurity measures and international cooperation to identify and mitigate the threats posed by such advanced persistent threat groups.
Cosa hanno in comune un mattoncino Lego e la backdoor XZ?Speck&Tech
ABSTRACT: A prima vista, un mattoncino Lego e la backdoor XZ potrebbero avere in comune il fatto di essere entrambi blocchi di costruzione, o dipendenze di progetti creativi e software. La realtà è che un mattoncino Lego e il caso della backdoor XZ hanno molto di più di tutto ciò in comune.
Partecipate alla presentazione per immergervi in una storia di interoperabilità, standard e formati aperti, per poi discutere del ruolo importante che i contributori hanno in una comunità open source sostenibile.
BIO: Sostenitrice del software libero e dei formati standard e aperti. È stata un membro attivo dei progetti Fedora e openSUSE e ha co-fondato l'Associazione LibreItalia dove è stata coinvolta in diversi eventi, migrazioni e formazione relativi a LibreOffice. In precedenza ha lavorato a migrazioni e corsi di formazione su LibreOffice per diverse amministrazioni pubbliche e privati. Da gennaio 2020 lavora in SUSE come Software Release Engineer per Uyuni e SUSE Manager e quando non segue la sua passione per i computer e per Geeko coltiva la sua curiosità per l'astronomia (da cui deriva il suo nickname deneb_alpha).
1. Revisiting Spatial-Temporal Similarity:
A Deep Learning Framework for Traffic Prediction
Association for the Advancement of Artificial Intelligence, 2019
Huaxiu Yao, Xianfeng Tang, Hua Wei, Guanjie Zheng, Zhenhui Li
January 22, 2021
Presenter: KyungHwan Moon
2. Contents
• Overview of the Paper
• Background and Motivation
• Introduction
• Spatial-Temporal Dynamic Network
• Experiment
• Conclusion and Discussion
3. 2
Overview of the Paper
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
• Local vs Non-Local
∎ Local : Using the size of the kernel during operation
(Local Operator)
∎ Non-Local : Using the non-local block for entire image
area(Non-Local Operator)
• Spatial & Temporal
• Static(Stationary) vs Dynamic
• Periodic & dependency
Xiaolong Wang, Ross Girshick, Abhinav Gupta, Kaiming He, Carnegie Mellon Universit, and Facebook AI Research.2018
4. 3
Background and Motivation
● Traffic prediction has drawn increasing attention in AI research field
∎ The increasing availability of large-scale traffic data and its importance in the real world
● Existing works make strong assumptions
∎ Spatial dependence is stationary in time
∎ Temporal dynamics is strictly periodical
● However, in practice the spatial dependence could be dynamic
∎ The spatial dependencies between locations are dynamic
∎ The temporal dependency follows daily and weekly pattern
but it is not strictly periodic for its dynamic temporal shifting
5. 4
Background and Motivation
• Previous methods
∎ Traffic time series for each individual location
(Li et al. 2012; Moreira-Matias et al. 2013;)
(Shekhar and Williams 2008; Lippi, Bertini, and Frasconi 2013)
• Recent studies
∎ Taking into account spatial information
-Adding regularizations on model similarity for nearby locations
(Deng etal. 2016; Id´eand Sugiyama 2011; Zheng and Ni 2013)
∎ Taking into account external context information
-Adding features of venue information, weather condition, and local events
(Wu, Wang, and Li 2016; Pan, Demiryurek, and Shahabi 2012; Tong et al. 2017)
※ Do not well capture the complex non-linear spatial-temporal dependency
6. 5
Background and Motivation
• Modeling the non-linear spatial dependency
∎ A heatmap image and Convolutional Neural Network(CNN)
(Zhang, Zheng, and Qi 2017; Zhang et al. 2016; Ma et al. 2017)
• Modeling the non-linear temporal dependency
∎ Recurrent Neural Network(RNN)-based framework
(Yuet al. 2017; Cui, Ke, and Wang 2016)
• Modeling both spatial and temporal dependencies
∎ Integrating CNN and Long Short-Term Memory(LSTM)
(Yao et al. 2018)
7. 6
Background and Motivation
• Limitation(1)
∎ The spatial dependency between locations relies only on the similarity of historical traffic
(Zhang, Zheng, and Qi 2017; Yao et al. 2018)
∎ The model learns a static spatial dependency
※ The dependencies between locations could change over time
• Limitation(2)
∎ Many existing studies ignore the shifting of long-term periodic dependency
∎ Traffic data show a strong daily and weekly periodicity and the dependency based on
such periodicity can be useful for prediction
※ The traffic data are not strictly periodic
※ Consider the sequential dependency and the temporal shifting in the periodicity
8. 7
Introduction
• Spatial-Temporal Dynamic Network (STDN)
∎ Based on a spatial-temporal neural network. which handles spatial and temporal
information via local CNN and LSTM, respectively
A flow-gated
local CNN
A periodically
shifted
attention
mechanism
⦁Spatial dependency by modeling the dynamic similarity
among locations using traffic flow informationation
⦁Learn the long-term periodic dependency and captures
both long-term periodic information and temporal shifting
in traffic sequence via attention mechanism
LSTM to handle the sequential dependency in a hierarchical way
10. 9
Introduction
• Datasets
∎ Taxi data of New York City(NYC)
∎ Bike-sharing data of NYC
• Contribution Summarization
∎ A flow gating mechanism to explicitly model dynamic spatial similarity
-The gate controls information propagation among nearby locations
∎ A peridically shifted attention mechanism by taking long-term periodic information
and temporal shifting simultaneously
∎ Experiments on several real-world traffic datasets
-The results show that our model is consistently better than other state-of-the-art
methods
12. 11
Spatial-Temporal Dynamic Network (STDN)
Periodically shifted attention mechanism (PSAM) captures the
long-term periodic dependency and temporal shifting. For each
day, we also use LSTM to capture the sequential information.
Figure 1: The architecture of STDN
(a)
13. 12
Spatial-Temporal Dynamic Network (STDN)
The short-term temporal dependency is captured by one LSTM.
Figure 1: The architecture of STDN
(b)
14. 13
Spatial-Temporal Dynamic Network (STDN)
The flow gating mechanism (FGM) tracks the dynamic spatial
similarity representation by controlling the spatial information
propagation; FC means fully connected layers and Conv means
several convolutional layers.
Figure 1: The architecture of STDN
(c)
15. 14
Spatial-Temporal Dynamic Network (STDN)
A unified multi-task prediction component predicts two types of traffic
volumes (start volume and end traffic volume) simultaneously.
Figure 1: The architecture of STDN
(d)
16. 15
Spatial-Temporal Dynamic Network (STDN)
• Local Spatial-Temporal Network
∎ Combine local CNN and LSTM (Yaoet al. 2018)
-To capture spatial and temporal squential dependency
-To deal with spatial and short-term temporal dependency
∎ Integrate and predict start and End volumes
-To mutually reinforce the prediction of two types of
traffic volumes
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
17. 16
Spatial-Temporal Dynamic Network (STDN)
• Local spatial dependency
∎ Convolutional neural network (CNN)
-To capture the spatial interactions
-treating the entire city as an image and simply applying CNN
may not achieve the best performance. Including regions
with weak correlations to predict a target region actually
hurts the performance.
∎ Use the local CNN
-To model the spatial dependency
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
For each time interval t, The target region i in the center of the image,
K convolutional layers
18. 17
Spatial-Temporal Dynamic Network (STDN)
• Short-term Temporal Dependency
∎ Long Short-Term Memory (LSTM)
-To address the exploding and vanishing gradient issue of
traditional Recurrent Neural Network (RNN)
-To capture the temporal sequential dependency
-The original version of LSTM
(Hochreiter and Schmidhuber 1997)
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
19. 18
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(1)
∎ CNN vs Local CNN
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
CNN Local CNN
⦁Handles the local structure similarity by
local connection and weight sharingion
⦁The local spatial dependency relies on
the similarity of historical traffic volume
⦁The spatial dependency of volume is
stationary, which can not fully reflect the
relation between the target region and its
neighbors.
n
Traffic flow is a more direct way to represent interactions between regions
20. 19
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(1)
∎ CNN vs Local CNN
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
CNN Local CNN
⦁Handles the local structure similarity by
local connection and weight sharingion
⦁The local spatial dependency relies on
the similarity of historical traffic volume
⦁The spatial dependency of volume is
stationary, which can not fully reflect the
relation between the target region and its
neighbors.
n
Traffic flow is a more direct way to represent interactions between regions
The relation between two regions is
stronger(i.e., they are more similar) if
there are more flows existing between
them
21. 20
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(2)
∎ Flow Gating Mechanism (FGM)
-To explicitly capture dynamic spatial dependency in the
hierarchy
-Construct the local spatial flow image to protect the
spatial to protect the spatial dependency of flow
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
Inflow Outflow
⦁Inflow departing from other location
ending in the region during the the time
intervalon
⦁Outflow starting from this region toward
somewhereon
The Traffic flow in two categories
22. 21
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(2)
∎ Flow Gating Mechanism (FGM)
-To explicitly capture dynamic spatial dependency in the
hierarchy
-Construct the local spatial flow image to protect the
spatial to protect the spatial dependency of flow
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
Inflow Outflow
⦁Inflow departing from other location
ending in the region during the the time
intervalon
⦁Outflow starting from this region toward
somewhereon
The Traffic flow in two categories
23. 22
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(3)
∎ Flow Gating Mechanism (FGM)
-The acquired flow matrices
-Use CNN to model the spatial flow interactions between
regions
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
24. 23
Spatial-Temporal Dynamic Network (STDN)
• Spatial Dynamic Similarity: Flow Gating Mechanism-(4)
∎ Flow Gating Mechanism (FGM)
-Use CNN to model the spatial flow interactions between
regions
After K gated convolutional layers, we use a flatten layer
followed by a fully connected layer to get the flow
gated spatial representation as
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
25. 24
Spatial-Temporal Dynamic Network (STDN)
• Temporal Dynamic Similarity
: Periodically Shifted Attention Mechanism (PSAM)-(1)
∎ Flow Gating Mechanism (FGM) overlooks the long-term
dependency (e.g. periodicity)which is an important
property of spatial-temporal prediction problem
(Zonoozi et al. 2018; Feng et al. 2018)
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
⦁Temoral shifting between different dayson ⦁Temoral shifting between different weeksn
The temporal shifting of periodicity
Each time in these figures represents a time interval (e.g., 9:30am means 9:00-9:30am)
26. 25
Spatial-Temporal Dynamic Network (STDN)
• Temporal Dynamic Similarity
: Periodically Shifted Attention Mechanism (PSAM)-(2)
∎ Training LSTM to handle long-term information is a
nontrivial task, since the increasing length enlarges the
risk of gradient vanishing, thus significantly weaken the
effects of periodicity.
∎ Periodically Shifted Attention Mechanism (PSAM)
-To take long-term periodic information into
consideration
-To tackle the limitations that the periodicity is not strict
daily or weekly
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
27. 26
Spatial-Temporal Dynamic Network (STDN)
• Temporal Dynamic Similarity
: Periodically Shifted Attention Mechanism (PSAM)-(2)
∎ Select Q time intervals from each day in Q to tackle the
potential temporal shifting
-For example, if the predicted time is 9:00-9:30pm, we
select 1 hour before and after the predicted time
(i.e., 8:00-10:30pm and |Q| = 5).
∎ Use LSTM to protect the sequential information
for each day
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
28. 27
Spatial-Temporal Dynamic Network (STDN)
• Temporal Dynamic Similarity
: Periodically Shifted Attention Mechanism (PSAM)-(3)
∎ Adopt an attention mechanism to capture the temporal
shifting and get the weighted representation of each
previous day.
-the representation of each previous days
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
29. 28
Spatial-Temporal Dynamic Network (STDN)
• Temporal Dynamic Similarity
: Periodically Shifted Attention Mechanism (PSAM)-(4)
∎ Adopt an attention mechanism to capture the temporal
shifting and get the weighted representation of each
previous day.
-the representation of each previous days
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
30. 29
Spatial-Temporal Dynamic Network (STDN)
• Joint Training-(1)
∎ Concatenate the short-term representation
and long-term representation
∎ For a fully connected layer, get the final prediction
value of start and end traffic volume for each region
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
31. 30
Spatial-Temporal Dynamic Network (STDN)
• Joint Training-(2)
∎ Predict start volume and end traffic volume
simultaneously
- Loss function
Local Spatial-Temporal Network
Local spatial dependency
Short-term Temporal Dependency
Spatial Dynamic Similarity:
Flow Gating Mechanism
Temporal Dynamic Similarity:
Periodically Shifted
Attention Mechanism
Joint Training
33. 32
Experiment
• Results
∎ Effectiveness of Flow Gating Mechanism (FGM)
-LSTN : Only short-term temporal dependency, and
local spatial dependency
-LSTN-FI : Use traffic flow information as features
instead of using a flow gating mechanism
-LSTN-FGM : Represent the spatial dynamic similarity between
local neighborhoods by utilizing flow gating
mechanism and do not use periodically shifted
attention mechanism
Performance
Comparison
Effectiveness of
Flow Gating Mechanism
Effectiveness of
Periodically Shifted
Attention Mechanism
34. 33
Experiment
• Results
∎ Evaluation of Flow Gating Mechanism (FGM)
-Root Mean Square Error (RMSE)
-Mean Absolute Percentage Error (MAPE)
Performance
Comparison
Effectiveness of
Flow Gating Mechanism
Effectiveness of
Periodically Shifted
Attention Mechanism
35. 34
Experiment
• Results
∎ Effectiveness of Periodically shifted attention Mechanism (PSAM)
-LSTN-L : Take long-term sequential information into
consideration to extend LSTN
-LSTN-SL : Remove the periodically shifted attention
mechanism in STDN and does not include flow gating
mechanism
-LSTN-PSAM : Add the periodically shifted attention mechanism
on LSTN-SL, only removing the flow gating
mechanism
Performance
Comparison
Effectiveness of
Flow Gating Mechanism
Effectiveness of
Periodically Shifted
Attention Mechanism
36. 35
Experiment
• Results
∎ Evaluation of Periodically shifted attention Mechanism
(PSAM)
-Root Mean Square Error (RMSE)
-Mean Absolute Percentage Error (MAPE)
Performance
Comparison
Effectiveness of
Flow Gating Mechanism
Effectiveness of
Periodically Shifted
Attention Mechanism
37. 36
Experiment
• Conclusion and Discussion
∎ Investigate the proposed model on other spatial-temporal prediction problems
∎ Explain the model
(i.e., explain feature importance of traffic prediction)
∎ Selecting appropriate hyperparameter in correspondence with data
∎ How much performance depending on the amount of data