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

Discovering temporal lagged and inter-dependencies in multivariate timeseries data is an important task. However, in many real-world applications,such as commercial cloud management, manufacturing predictive maintenance, andportfolios performance analysis, such dependencies can be non-linear andtime-variant, which makes it more challenging to extract such dependenciesthrough traditional methods such as Granger causality or clustering. In thiswork, we present a novel deep learning model that uses multiple layers ofcustomized gated recurrent units (GRUs) for discovering both time laggedbehaviors as well as inter-timeseries dependencies in the form of directedweighted graphs. We introduce a key component of Dual-purpose recurrent neuralnetwork that decodes information in the temporal domain to discover laggeddependencies within each time series, and encodes them into a set of vectorswhich, collected from all component time series, form the informative inputs todiscover inter-dependencies. Though the discovery of two types of dependenciesare separated at different hierarchical levels, they are tightly connected andjointly trained in an end-to-end manner. With this joint training, learning ofone type of dependency immediately impacts the learning of the other one,leading to overall accurate dependencies discovery. We empirically test ourmodel on synthetic time series data in which the exact form of (non-linear)dependencies is known. We also evaluate its performance on two real-worldapplications, (i) performance monitoring data from a commercial cloud provider,which exhibit highly dynamic, non-linear, and volatile behavior and, (ii)sensor data from a manufacturing plant. We further show how our approach isable to capture these dependency behaviors via intuitive and interpretabledependency graphs and use them to generate highly accurate forecasts.

seq2graph: Discovering Dynamic Dependencies from Multivariate Time Series with Multi-level Attention