Statistical inference for noisy nonlinear ecological dynamic systems

By Simon N. Wood. 2010

640 citations recorded by the source corpus.

Subfields: Ecology, Global and Planetary Change, Nature and Landscape Conservation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecology

In Global and Planetary Change

In Nature and Landscape Conservation

Citations in the corpus

Gradient-free Hamiltonian Monte Carlo with Efficient Kernel Exponential Families (2015, methodology): “Accordingly, Hamiltonian ABC was recently proposed [8], combining the synthetic likelihood approach [26] with gradients based on stochastic finite differences.”
K2-ABC: Approximate Bayesian Computation with Infinite Dimensional Summary Statistics via Kernel Embeddings (2015, methodology): “As an example of statistical inference for ecological dynamic systems, we use observations on adult blowfly populations over time introduced in Wood (2010).”
Variational Approaches for Auto-Encoding Generative Adversarial Networks (2017, methodology): “The original formulation of the synthetic likelihood [42] is based on a Gaussian assumption, but we use the term here to mean any general substitute for the likelihood that maintains its asymptotic properties.”
Distribution Matching in Variational Inference (2018, background): “This approach is a form of synthetic likelihood (Dutta et al., 2016, Wood, 2010).”
Bayesian Synthetic Likelihood (2017, methodology): “We consider the Ricker model presented in Wood (2010). Here a population of size Nt at time t evolves according to Nt+1 = rNte −Nt+et where et iid ∼ N (0, σ(2) e).”
Likelihood-free inference with emulator networks (2018, methodology): “The likelihoods of the emulator can then be plugged into any Bayesian inference approach (as in synthetic likelihood approaches [10, 11, 36]) to calculate the posterior.”
BayesFlow: Learning Complex Stochastic Models With Invertible Neural Networks (2020, methodology): “Likelihood-free problems arise, for example, when p(x |θ) is not available in closed-form, or when the forward process is defined by a stochastic differential equation, a computer simulation, or a complicated algorithm [30, 52, 51, 56].”
Flexible statistical inference for mechanistic models of neural dynamics (2017, methodology): “Synthetic-likelihood methods [28, 21, 29] approximate the likelihood p(x|θ) using multivariate Gaussians fitted to repeated simulations given θ (see [30, 31] for generalisations).”
Statistical inference for dynamical systems: A review (2012, background): “Let us mention that very often both dynamical noise and observational noise arise in real data examples, especially in biology and ecology [15, 35, 91, 96, 107, 164, 173]; nonetheless, as the other settings arise in some physical models, we discuss each of the logical possibilities for the noise str”
Asymptotically exact inference in differentiable generative models (2016, methodology): “In this case conditioning on the exact observed data could be viewed as over constraining the model to reproduce features of the data which are only incidental, and using summary statistics which are invariant to phase could be preferable to conditioning on the full data [99].”
Dynamic Likelihood-free Inference via Ratio Estimation (DIRE) (2018, methodology): “Related (nonlinear) autocorrelation functions have been used as summary statistics in previous work (e.g. Wood 2010).”
Likelihood-Free Inference by Ratio Estimation (2016, methodology): “, 2002) and the synthetic likelihood approach (Wood, 2010).”
On the Asymptotic Efficiency of Approximate Bayesian Computation Estimators (2015, methodology): “There is a wide class of such likelihood-free methods of inference including indirect inference (Gouriéroux & Ronchetti, 1993; Heggland & Frigessi, 2004), the bootstrap filter (Gordon et al., 1993), simulated methods of moments (Duffie & Singleton, 1993), and synthetic likelihood (Wood, 2010).”
Bayesian Prediction of Future Street Scenes using Synthetic Likelihoods (2018, methodology): “We address this problem through an objective leveraging synthetic likelihoods (Wood, 2010; Rosca et al., 2017) which relaxes the constraint on every model to explain every data point, thus encouraging diversity in the learned models to deal with multi-modality.”
A Review on Deep Learning Techniques for Video Prediction (2020, background): “As main contribution, their dropout-based Bayesian approach leverages synthetic likelihoods [179], [180] to encourage prediction diversity and deal with multi-modal outcomes.”
Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flows (2018, methodology): “Synthetic Likelihood [20, 57, 63, 85] estimates the mean m θ and co-variance matrix S θ of a batch of data x n ∼ p ( x | θ ) sampled at a given θ , and then approximates p ( x | ≈ Non-Gaussian likelihood approximations are also possible, e.g. saddlepoint approximations [22].”
Approximate Bayesian computation via regression density estimation (2012, background): “Wood (2010) considers a “synthetic likelihood” which involves estimating a normal distribution for the summary statistics, S, with mean and covariance depending on θ.”
GPS-ABC: Gaussian Process Surrogate Approximate Bayesian Computation (2014, methodology): “We next discuss the approximation introduced by Wood [28], who models the simulated pseudo-data {x1, ..,xS} at parameter value θ using a normal distribution, i.e. π(x|θ) ≈ N (x|µ̂θ, Σ̂θ).”