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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Generalization of time series prediction remains an important open issue inmachine learning, wherein earlier methods have either large generalizationerror or local minima. We develop an analytically solvable, unsupervisedlearning scheme that extracts the most informative components for predictingfuture inputs, termed predictive principal component analysis (PredPCA). Ourscheme can effectively remove unpredictable noise and minimize test predictionerror through convex optimization. Mathematical analyses demonstrate that,provided with sufficient training samples and sufficiently high-dimensionalobservations, PredPCA can asymptotically identify hidden states, systemparameters, and dimensionalities of canonical nonlinear generative processes,with a global convergence guarantee. We demonstrate the performance of PredPCAusing sequential visual inputs comprising hand-digits, rotating 3D objects, andnatural scenes. It reliably estimates distinct hidden states and predictsfuture outcomes of previously unseen test input data, based exclusively onnoisy observations. The simple architecture and low computational cost ofPredPCA are highly desirable for neuromorphic hardware.

Dimensionality reduction to maximize prediction generalization capability