Approximate Inference for Multiplicative Latent Force Models
Authors:
Daniel J. Tait,
Bruce J. Worton
Abstract:
Latent force models are a class of hybrid models for dynamic systems, combining simple mechanistic models with flexible Gaussian process (GP) perturbations. An extension of this framework to include multiplicative interactions between the state and GP terms allows strong a priori control of the model geometry at the expense of tractable inference.
In this paper we consider two methods of carrying…
▽ More
Latent force models are a class of hybrid models for dynamic systems, combining simple mechanistic models with flexible Gaussian process (GP) perturbations. An extension of this framework to include multiplicative interactions between the state and GP terms allows strong a priori control of the model geometry at the expense of tractable inference.
In this paper we consider two methods of carrying out inference within this broader class of models. The first is based on an adaptive gradient matching approximation, and the second is constructed around mixtures of local approximations to the solution. We compare the performance of both methods on simulated data, and also demonstrate an application of the multiplicative latent force model on motion capture data.
△ Less
Submitted 31 December, 2018;
originally announced December 2018.
Multiplicative Latent Force Models
Authors:
Daniel J. Tait,
Bruce J. Worton
Abstract:
Bayesian modelling of dynamic systems must achieve a compromise between providing a complete mechanistic specification of the process while retaining the flexibility to handle those situations in which data is sparse relative to model complexity, or a full specification is hard to motivate. Latent force models achieve this dual aim by specifying a parsimonious linear evolution equation which an ad…
▽ More
Bayesian modelling of dynamic systems must achieve a compromise between providing a complete mechanistic specification of the process while retaining the flexibility to handle those situations in which data is sparse relative to model complexity, or a full specification is hard to motivate. Latent force models achieve this dual aim by specifying a parsimonious linear evolution equation which an additive latent Gaussian process (GP) forcing term. In this work we extend the latent force framework to allow for multiplicative interactions between the GP and the latent states leading to more control over the geometry of the trajectories. Unfortunately inference is no longer straightforward and so we introduce an approximation based on the method of successive approximations and examine its performance using a simulation study.
△ Less
Submitted 1 November, 2018;
originally announced November 2018.
Point process models for quasi-periodic volcanic earthquakes
Authors:
Anastasia Ignatieva,
Andrew F. Bell,
Bruce J. Worton
Abstract:
Long period (LP) earthquakes are common at active volcanoes, and are ubiquitous at persistently active andesitic and dacitic subduction zone volcanoes. They provide critical information regarding the state of volcanic unrest, and their occurrence rates are key data for eruption forecasting. LPs are commonly quasi-periodic or 'anti-clustered', unlike volcano-tectonic (VT) earthquakes, so the existi…
▽ More
Long period (LP) earthquakes are common at active volcanoes, and are ubiquitous at persistently active andesitic and dacitic subduction zone volcanoes. They provide critical information regarding the state of volcanic unrest, and their occurrence rates are key data for eruption forecasting. LPs are commonly quasi-periodic or 'anti-clustered', unlike volcano-tectonic (VT) earthquakes, so the existing Poisson point process methods used to model occurrence rates of VT earthquakes are unlikely to be optimal for LP data. We evaluate the performance of candidate formulations for LP data, based on inhomogeneous point process models with four different inter-event time distributions: exponential (IP), Gamma (IG), inverse Gaussian (IIG), and Weibull (IW). We examine how well these models explain the observed data, and the quality of retrospective forecasts of eruption time. We use a Bayesian MCMC approach to fit the models. Goodness-of-fit is assessed using Quantile-Quantile and Kolmogorov-Smirnov methods, and benchmarking against results obtained from synthetic datasets. IG and IIG models were both found to fit the data well, with the IIG model slightly outperforming the IG model. Retrospective forecasting analysis shows that the IG model performs best, with the initial preference for the IIG model controlled by catalogue incompleteness late in the sequence. The IG model fits the data significantly better than the IP model, and simulations show it produces better forecasts for highly periodic data. Simulations also show that forecast precision increases with the degree of periodicity of the earthquake process using the IG model, and so should be better for LP earthquakes than VTs. These results provide a new framework for point process modelling of volcanic earthquake time series, and verification of alternative models.
△ Less
Submitted 2 September, 2018; v1 submitted 20 March, 2018;
originally announced March 2018.
