English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Computer simulations have proven a valuable tool for understanding complexphenomena across the sciences. However, the utility of simulators for modellingand forecasting purposes is often restricted by low data quality, as well aspractical limits to model fidelity. In order to circumvent these difficulties,we argue that modellers must treat simulators as idealistic representations ofthe true data generating process, and consequently should thoughtfully considerthe risk of model misspecification. In this work we revisit neural posteriorestimation (NPE), a class of algorithms that enable black-box parameterinference in simulation models, and consider the implication of asimulation-to-reality gap. While recent works have demonstrated reliableperformance of these methods, the analyses have been performed using syntheticdata generated by the simulator model itself, and have therefore only addressedthe well-specified case. In this paper, we find that the presence ofmisspecification, in contrast, leads to unreliable inference when NPE is usednaively. As a remedy we argue that principled scientific inquiry withsimulators should incorporate a model criticism component, to facilitateinterpretable identification of misspecification and a robust inferencecomponent, to fit 'wrong but useful' models. We propose robust neural posteriorestimation (RNPE), an extension of NPE to simultaneously achieve both theseaims, through explicitly modelling the discrepancies between simulations andthe observed data. We assess the approach on a range of artificiallymisspecified examples, and find RNPE performs well across the tasks, whereasnaively using NPE leads to misleading and erratic posteriors.

Robust Neural Posterior Estimation and Statistical Model Criticism