Variational Inference for Infinite Mixtures of Gaussian Processes With Applications to Traffic Flow Prediction

Sun, Shiliang; Xu, Xiaojun

doi:10.1109/tits.2010.2093575

Cited by 126 publications

(59 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Traffic flow prediction is a problem which has been intensively studied for a long time [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33]. Both the traffic on highways and the traffic in city networks is considered.…”

Section: State Of the Artmentioning

confidence: 99%

Prediction of Traffic Intensity for Dynamic Street Lighting

Bielecka

Bielecki

Ernst

et al. 2017

Proceedings of the 2017 Federated Conference on Computer Science and Information Systems

View full text Add to dashboard Cite

Abstract-In this paper, the problem of short-term prediction of traffic flow in a city traffic network is considered. This prediction is performed in order to provide input data to a dynamic control system for street lighting. The forecasting is done by a multi-layer using artificial neural network. Because of the limited number of sensors, the data is insufficient to describe the relation between the traffic intensity at a given point and the points in which the flow intensity is measured. The proposed approach is tested by using data from the centre of Kraków. The prediction error turned to be low.

show abstract

Section: State Of the Artmentioning

confidence: 99%

Prediction of Traffic Intensity for Dynamic Street Lighting

Bielecka

Bielecki

Ernst

et al. 2017

Proceedings of the 2017 Federated Conference on Computer Science and Information Systems

View full text Add to dashboard Cite

show abstract

“…Therefore, following [17,19,22], we avoid this problem by using an infinite number of experts for our model. In particular, a Dirichlet process (DP) prior [27] is placed over the experts to allow the model to automatically determine the number of components.…”

Section: Dirichlet Process Mixture Modelmentioning

confidence: 99%

“…Recently, several variational mixtures of GP experts [20,22] have been proposed to use variational inference [23], which is a more flexible and faster alternative to MCMC sampling. In these methods, each GP expert has its own set of inducing points, and is described by a linear model.…”

Section: Introductionmentioning

confidence: 99%

“…In our proposed method, each GP expert accompanied by its own set of inducing points is a sparse approximation to the full GP. Instead of imposing a deterministic relation between latent variables at training points and those at inducing points like in [20,22], our model uses an approximate conditional distribution of the latent variables assuming that they are independent given the inducing points. In this way, variational inference in our model is made feasible without suffering from the degeneracy problem mentioned above.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Variational inference for infinite mixtures of sparse Gaussian processes through KL-correction

Nguyen

Bouzerdoum

Phung

2016

2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

We propose a new approximation method for Gaussian process (GP) regression based on the mixture of experts structure and variational inference. Our model is essentially an infinite mixture model in which each component is composed of a Gaussian distribution over the input space, and a Gaussian process expert over the output space. Each expert is a sparse GP model augmented with its own set of inducing points. Variational inference is made feasible by assuming that the training outputs are independent given the inducing points. In previous works on variational mixture of GP experts, the inducing points are selected through a greedy selection algorithm, which is computationally expensive. In our method, both the inducing points and hyperparameters of the experts are learned through maximizing an improved lower bound of the marginal likelihood. Experiments on benchmark datasets show the advantages of the proposed method. VARIATIONAL INFERENCE FOR INFINITE MIXTURES OF SPARSE GAUSSIAN PROCESSES THROUGH KL-CORRECTION T. N. A. Nguyen, A. Bouzerdoum, and S. L. PhungSchool of Electrical, Computer and Telecommunications Engineering, University of Wollongong, NSW, 2522, Australia ABSTRACTWe propose a new approximation method for Gaussian process (GP) regression based on the mixture of experts structure and variational inference. Our model is essentially an infinite mixture model in which each component is composed of a Gaussian distribution over the input space, and a Gaussian process expert over the output space. Each expert is a sparse GP model augmented with its own set of inducing points. Variational inference is made feasible by assuming that the training outputs are independent given the inducing points.In previous works on variational mixture of GP experts, the inducing points are selected through a greedy selection algorithm, which is computationally expensive. In our method, both the inducing points and hyperparameters of the experts are learned through maximizing an improved lower bound of the marginal likelihood. Experiments on benchmark datasets show the advantages of the proposed method. Index Terms-Gaussian process, variational inference.

show abstract

“…If the probability of association of trajectory to the closest motion pattern is less than a threshold, the trajectory is treated as anomalous. Sun and Xu [31] employ infinite mixtures of Gaussian to model the trajectory patterns. They automatically identify the number of mixtures that are required to effectively model the given trajectory data.…”

Section: Introductionmentioning

confidence: 99%