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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The SUBNET neural network architecture has been developed to identifynonlinear state-space models from input-output data. To achieve this, itcombines the rolled-out nonlinear state-space equations and a state encoderfunction, both parameterised as neural networks The encoder function isintroduced to reconstruct the current state from past input-output data. Hence,it enables the forward simulation of the rolled-out state-space model. Whilethis approach has shown to provide high-accuracy and consistent modelestimation, its convergence can be significantly improved by efficientinitialization of the training process. This paper focuses on such aninitialisation of the subspace encoder approach using the Best LinearApproximation (BLA). Using the BLA provided state-space matrices and itsassociated reconstructability map, both the state-transition part of thenetwork and the encoder are initialized. The performance of the improvedinitialisation scheme is evaluated on a Wiener-Hammerstein simulation exampleand a benchmark dataset. The results show that for a weakly nonlinear system,the proposed initialisation based on the linear reconstructability map resultsin a faster convergence and a better model quality.

Initialization Approach for Nonlinear State-Space Identification via the Subspace Encoder Approach