Skip to main content

Cortex Features

Cortex processes mindLAMP data through three tiers. Raw features fetch sensor and activity data from the LAMP API into a consistent format. Primary features are sparse, event-based — they identify discrete events or periods (e.g., individual trips, screen-on bouts, location clusters) with variable counts per time range. Secondary features are dense, time-windowed — they aggregate data into exactly one value per fixed time window (controlled by the resolution parameter, default 1 day), producing regular time-series suitable for statistical analysis.

Feature Tiers​

Raw Features
Raw feature modules for direct sensor and activity data.
Primary Features
Primary feature modules with output schemas.
Secondary Features
Secondary feature modules organized by category.

Pipeline Overview​

Data SourcesRawPrimarySecondarylamp.gpsgpssignificant_locationstripshometimeentropytrip_distancetrip_durationdata_qualitylamp.accelerometeraccelerometeracc_jerkinactive_durationlamp.device_statedevice_statescreen_activescreen_durationbattery_levellamp.stepsstepsstep_countlamp.nearby_devicenearby_devicenearby_device_countlamp.telephonytelephonycall_numbercall_durationcall_degreelamp.sleepsleephealthkit_sleep_durationlamp.surveysurveysurvey_scoressurvey_resultslamp.jewels_alamp.jewels_blamp.balloon_risklamp.cats_and_dogslamp.pop_the_bubbleslamp.spatial_spanjewels_ajewels_bballoon_riskcats_and_dogspop_the_bubblesspatial_spangame_level_scoresgame_resultslamp.analyticsanalyticsfeeds step_count, app_timeSENSORKIT (iOS RESEARCH ENROLLMENT)device_usagescreen_unlocksscreen_unlock_durationscreen_wakesapp_timemessages_usagetext_numbertext_degreevisitsvisit_timeFull pipeline (raw → primary → secondary)Direct (raw → secondary, skipping primary)

Raw features virtualize low-level data streams into a consistent format, handling iOS/Android harmonization automatically. Primary features produce sparse output — a variable number of discrete events per time range (e.g., a list of trips or screen-on bouts, each with start/end timestamps). Secondary features produce dense output — exactly one value per time window (e.g., total daily screen time, daily step count), making them directly suitable for statistical analysis and clinical reporting.

Cortex automatically manages dependencies — requesting a secondary feature like hometime triggers computation of all required primary and raw features. Raw features handle iOS/Android harmonization automatically, so downstream features receive uniform data regardless of the source platform.

SensorKit features

Some features (screen_unlocks, text_number, visit_time, etc.) depend on iOS SensorKit data, available only on devices enrolled in Apple-approved research studies. Most deployments use the standard sensor equivalents instead. See the Secondary Features page for details.