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

We adapt image inpainting techniques to impute large, irregular missingregions in urban settings characterized by sparsity, variance in both space andtime, and anomalous events. Missing regions in urban data can be caused bysensor or software failures, data quality issues, interference from weatherevents, incomplete data collection, or varying data use regulations; anymissing data can render the entire dataset unusable for downstreamapplications. To ensure coverage and utility, we adapt computer visiontechniques for image inpainting to operate on 3D histograms (2D space + 1Dtime) commonly used for data exchange in urban settings.  Adapting these techniques to the spatiotemporal setting requires handlingskew: urban data tend to follow population density patterns (small denseregions surrounded by large sparse areas); these patterns can dominate thelearning process and fool the model into ignoring local or transient effects.To combat skew, we 1) train simultaneously in space and time, and 2) focusattention on dense regions by biasing the masks used for training to the skewin the data. We evaluate the core model and these two extensions using the NYCtaxi data and the NYC bikeshare data, simulating different conditions formissing data. We show that the core model is effective qualitatively andquantitatively, and that biased masking during training reduces error in avariety of scenarios. We also articulate a tradeoff in varying the number oftimesteps per training sample: too few timesteps and the model ignorestransient events; too many timesteps and the model is slow to train withlimited performance gain.

Adapting to Skew: Imputing Spatiotemporal Urban Data with 3D Partial Convolutions and Biased Masking