• Corpus ID: 60441440

Bayesian Online Detection and Prediction of Change Points

@article{AgudeloEspaa2019BayesianOD,
  title={Bayesian Online Detection and Prediction of Change Points},
  author={Diego Agudelo-Espa{\~n}a and Sebasti{\'a}n G{\'o}mez-Gonz{\'a}lez and Stefan Bauer and Bernhard Sch{\"o}lkopf and Jan Peters},
  journal={ArXiv},
  year={2019},
  volume={abs/1902.04524}
}
Online detection of instantaneous changes in the generative process of a data sequence generally focuses on retrospective inference of such change points without considering their future occurrences. We extend the Bayesian Online Change Point Detection algorithm to also infer the number of time steps until the next change point (i.e., the residual time). This enables us to handle observation models which depend on the total segment duration, which is useful to model data sequences with temporal… 

Figures and Tables from this paper

Bayesian Online Prediction of Change Points

TLDR
The Bayesian Online Change Point Detection algorithm is extended to also infer the number of time steps until the next change point (i.e., the residual time), which enables to handle observation models which depend on the total segment duration, which is useful to model data sequences with temporal scaling.

Bayesian Online Change Point Detection for Baseline Shifts

  • Ginga Yoshizawa
  • Computer Science
    Statistics, Optimization & Information Computing
  • 2020
TLDR
The original BOCPD algorithm is extended to be applicable to a time series whose baseline is constantly shifting toward unknown values but also why the proposed extension works is visualize to demonstrate the efficacy of the proposed algorithm.

Change-Point Detection on Hierarchical Circadian Models

Deep Explicit Duration Switching Models for Time Series

TLDR
The Recurrent Explicit Duration Switching Dynamical System (RED-SDS), a model capable of identifying both state-and time-dependent switching dynamics, is proposed that achieves considerable improvement in time series segmentation and competitive forecasting performance against the state of the art.

Real Time Probabilistic Models for Robot Trajectories

TLDR
A learning method to learn a movement primitive from a small set of demonstrations performed by a human expert is proposed, and a new method for forecasting the future value of a trajectory given its past observations based on variational auto-encoders is introduced.

References

SHOWING 1-10 OF 36 REFERENCES

Bayesian Online Changepoint Detection

TLDR
This work examines the case where the model parameters before and after the changepoint are independent and derives an online algorithm for exact inference of the most recent changepoint and its implementation is highly modular so that the algorithm may be applied to a variety of types of data.

A survey of methods for time series change point detection

TLDR
This survey article enumerates, categorizes, and compares many of the methods that have been proposed to detect change points in time series, and presents some grand challenges for the community to consider.

Gaussian Process Change Point Models

TLDR
A nonparametric time series model which can handle change points and, unlike other Bayesian online change point detection algorithms, is applicable when temporal correlations in a regime are expected.

Bayesian Retrospective Multiple‐Changepoint Identification

TLDR
This work discusses the use of a sampling‐based technique, the Gibbs sampler, in multiple‐changepoint problems and demonstrates how it can be used to reduce the computational load involved considerably.

A review of change point detection methods

On‐line inference for multiple changepoint problems

TLDR
It is shown how resampling ideas from particle filters can be used to reduce the computational cost to linear in the number of observations, at the expense of introducing small errors, and two new, optimum resamplings algorithms are proposed for this problem.

Structured Inference for Recurrent Hidden Semi-markov Model

TLDR
A structured and stochastic sequential neural network (SSNN), which composes with a generative network and an inference network that aims to not only capture the long-term dependencies but also model the uncertainty of the segmentation labels via semi-Markov models.

Learning and Inferring Motion Patterns using Parametric Segmental Switching Linear Dynamic Systems

TLDR
The overall development of the extensions for SLDSs provide a principled framework to interpret complex motions by incorporating additional global parameters that underly systematic variations of the overall motion.

Recurrent Hidden Semi-Markov Model

TLDR
The Recurrent Neural Network (RNN) is proposed to incorporate the Recurrent HSMM as the generative process of each segment, resulting the Recurrence HSMM (R-HSMM), which significantly outperforms the previous state-of-the-art on both the synthetic and real-world datasets.

An efficient forward-backward algorithm for an explicit-duration hidden Markov model

TLDR
A new forward-backward algorithm is proposed whose computational complexity is only O((MD + M/sup 2/)T), a reduction by almost a factor of D when D > M and whose memory requirement is O(MT).