Dataset Catalog¶

Thirteen ready-to-use spatio-temporal point-process datasets, hosted on the seahorse-stpp Hugging Face organization and loadable by a short name. They share one trait — events observed in space and time — but the meaning of "space" stretches from a city block to the human cortex. The catalog is ordered along that arc.

Every dataset loads the same way, by the short name shown on each card:

from seahorse.data import load_dataset

splits = load_dataset("citibike")  # {"train": [...], "val": [...], "test": [...]}

…or from the command line with --dataset citibike. The full seahorse-stpp/<id> reference still works everywhere a short name does. Three of them — marked core benchmark below — are the 2D-spatial trio used for the headline real-data comparison.

Moving through the city¶

space · streets, docks, and pickup points

NYC Citibikecore benchmark citibike Bike-share trips between docking stations across New York City. spacedock coordinatestimeride start Uber Pickups · NYC uber_pickups Ride-hailing pickup requests across New York City. spacepickup locationtimerequest time US Traffic Accidents us_accidents Road traffic accident reports across the United States. spaceaccident locationtimereport time

Incidents & safety¶

space · where something was reported

Chicago Crime chicago_crime Reported crime incidents across the city of Chicago. spaceblock locationtimeincident time LA Crime la_crime Reported crime incidents across Los Angeles. spacereport locationtimeincident time Global Terrorism Database gtd Worldwide terrorism events from the GTD. spaceevent locationtimeevent date Austin 311 austin_311 Non-emergency city service requests in Austin, Texas. spacerequest locationtimerequest time

Earth & environment¶

space · latitude and longitude on the globe

Earthquakescore benchmark earthquakes Seismic events with epicenter and origin time. spaceepicentertimeorigin time US Wildfires us_wildfires Wildfire ignition records across the United States. spaceignition locationtimediscovery date

Population health¶

space · case geography

COVID-19core benchmark covid Reported COVID-19 cases over space and time. spacecase locationtimereport date

space · where people check in

Gowalla gowalla Location-based social check-ins from the Gowalla network. spacevenue locationtimecheck-in time Brightkite brightkite Location-based social check-ins from the Brightkite network. spacevenue locationtimecheck-in time

Beyond the map¶

space · not a map at all

BOLD5000non-2D bold5000 Eventized neural responses derived from BOLD5000 fMRI recordings — included to show STPP space need not be geographic. spacethe braintimescan time

Supported — but not in the headline benchmark

BOLD5000 ships as a fully supported dataset. The main real-data comparison stays on the 2D-spatial trio — COVID, Earthquakes, and Citibike — because several benchmark presets are built for two-dimensional spatial event domains.

Synthetic benchmark suites¶

space · whatever you configure it to be

When you need known ground truth — to isolate one factor that real data confounds — Seahorse uses synthetic sequences from HawkesNest.

HawkesNest · entanglement suitesynthetic github.com/YahyaAalaila/HawkesNest Spatio-temporal sequences generated with tunable entanglement between the space and time dimensions, under known ground truth — so a model's behaviour can be probed against a factor that real-world data cannot cleanly separate. spaceconfigurable domaintimeconfigurable process

Want to add your own? See Add Your Dataset for the preparation checklist and Conversion Standard for the JSONL format.