Data Details

Each row in your shift data is a shift; the following are helpful descriptions of columns within that dataset: ● “Agent ID”: HCP ID ● “Facility ID”: HCF ID ● “Start”: The shift start time ● Agent Req”: the type of HCP that is being requested for this shift ● “End”: The shift end time ● “Shift Type”: specifies if the shift is in the morning (AM), afternoon (PM), overnight (NOC), or custom (CUSTOM) ● “Deleted”: Whether the shift was deleted ○ Note “deleted” means “canceled by facility” ● “Created At”: When the shift was created ● “Charge”: Per hour charge rate ● “Time”: How many hours the shift lasts ● “Verified”: Indicates that the shift was worked, as confirmed by a signed timesheet

Each row in your cancellation logs is a unique cancellation event; the following are helpful descriptions of columns within that dataset: ● “Action”: The type of cancellation action ○ “WORKER_CANCEL”: The HCP canceled a shift they booked ○ “NO_CALL_NO_SHOW”: The HCP canceled a shift they booked after the shift commenced or otherwise did not show up to the shift and did not inform the facility about their absence ● “Created At”: When the action took place ● “Facility ID”: HCF ID ● “Worker ID”: The ID of the HCP that was previously associated with the shift ● “Shift ID”: The shift ID ● “Lead Time”: The time from “action” to “shift start” (in hours)

Each row in your shift claim logs is a unique booking event; the following are helpful descriptions for columns within that dataset:

Note that we only included claim actions for a subset of the date range in the "shifts" data. Thus, there are likely shifts that don't have associated claim actions. That's OK, we're only providing this data so you can observe HCP booking behavior. ● “Action”: The type of booking action ○ "SHIFT_CLAIM": The HCP instantly booked the shift. As soon as they booked the shift, it was theirs.

Business Problem

You’ll likely want to know more about how the marketplace is currently operating to form your own mental model.

Data ● In this “Data” folder, you can find the below: ○ Shift data for one of the metropolitan statistical areas in which we have a presence ○ A list of cancellation logs for shifts that were canceled by HCPs ○ A list of shift claim logs ● We define the fields in these files below

Assumptions and Business Context ● The most damaging type of cancellation for the HCF is one in which the HCP does what we call a “No-Call-No-Show”; this means they canceled the shift after the shift started or otherwise did not show up to the shift and did not inform the facility of their absence ● The top reasons why HCPs cancel shifts last minute are: sick, family emergency, transportation issue (e.g. car broke down), facility issue ● From interviews, the most important things to HCPs are: will there be shifts that fit my erratic schedule, that are close enough to home, that pay enough, and that pay on time? ● HCPs currently receive a set of notifications prior to their shift to remind them of their upcoming shift

Benchmarking the robustness to distribution shifts traditionally relies on dataset collection which is typically laborious and expensive, in particular for datasets with a large number of classes like ImageNet. An exception to this procedure is ImageNet-C (Hendrycks & Dietterich, 2019), a dataset created by applying common real-world corruptions at different levels of intensity to the (clean) ImageNet images. Inspired by this work, we introduce ImageNet-Cartoon and ImageNet-Drawing, two datasets constructed by converting ImageNet images into cartoons and colored pencil drawings, using a GAN framework (Wang & Yu, 2020) and simple image processing (Lu et al., 2012), respectively.

This repository contains ImageNet-Cartoon and ImageNet-Drawing. Checkout the official GitHub Repo for the code on how to reproduce the datasets.

If you find this useful in your research, please consider citing:

@inproceedings{imagenetshift,
 title={ImageNet-Cartoon and ImageNet-Drawing: two domain shift datasets for ImageNet},
 author={Tiago Salvador and Adam M. Oberman},
 booktitle={ICML Workshop on Shift happens: Crowdsourcing metrics and test datasets beyond ImageNet.},
 year={2022}
}

AFSC triennial and NWFSC annual shelf-slope bottom trawl surveys (spatially-restricted)A subset of locations and species obtained from the Alaska Fisheries Science Center triennial shelf survey (1977-2004) and Northwest Fisheries Science Center annual shelf-slope survey (2003-2013). Please see ReadMe.txt for more detailsArchive.csv

This dataset provides calculated remote sensing reflectance (Rrs) from measurements collected with a Ramses TriOS radiometer deployed on the Santa Barbara Museum of Natural History Sea Center at Stearns Wharf, Santa Barbara, California, U.S. All measurements were taken over a fixed position at (34.41037665, -119.68557147). Three sensors are used to collect solar downwelling irradiance (Ed), sky radiance (Ls) and water-leaving radiance (Lw). These data have been processed to Rrs at 10 second intervals and are either concurrent or taken within 2.5 hours of SHIFT campaign flights. The data collected by the three Ramses TriOS sensors for eight days during the period 2022-04-05 to 2022-05-29 are also included. The data were translated from the proprietary format output by the Ramses TriOS instrument and saved in comma-separated values (CSV) format.

While the traditional viewpoint in machine learning and statistics assumes training and testing samples come from the same population, practice belies this fiction. One strategy—coming from robust statistics and optimization—is thus to build a model robust to distributional perturbations. In this paper, we take a different approach to describe procedures for robust predictive inference, where a model provides uncertainty estimates on its predictions rather than point predictions. We present a method that produces prediction sets (almost exactly) giving the right coverage level for any test distribution in an f-divergence ball around the training population. The method, based on conformal inference, achieves (nearly) valid coverage in finite samples, under only the condition that the training data be exchangeable. An essential component of our methodology is to estimate the amount of expected future data shift and build robustness to it; we develop estimators and prove their consistency for protection and validity of uncertainty estimates under shifts. By experimenting on several large-scale benchmark datasets, including Recht et al.’s CIFAR-v4 and ImageNet-V2 datasets, we provide complementary empirical results that highlight the importance of robust predictive validity.

This dataset contains the predicted prices of the asset Shift AI over the next 16 years. This data is calculated initially using a default 5 percent annual growth rate, and after page load, it features a sliding scale component where the user can then further adjust the growth rate to their own positive or negative projections. The maximum positive adjustable growth rate is 100 percent, and the minimum adjustable growth rate is -100 percent.

Shift is a lightweight framework for high performance local and remote file transfers that provides resiliency across a wide variety of failure scenarios through various techniques. These include end-to-end integrity via cryptographic hashes, throttling of transfers to prevent resource exhaustion, balancing transfers across resources based on load and availability, and parallelization of transfers across multiple source and destination hosts for increased redundancy and performance.

This archive contains part 2 of Shift Benchmark on Multiple Sclerosis lesion segmentation data. This dataset is provided by the Shifts Project to enable assessment of the robustness of models to distributional shift and the quality of their uncertainty estimates. This part is contains data collected from several different sources and distributed under a CC BY NC SA 4.0 license. Part 1 of the data is available here. A full description of the benchmark is available in https://arxiv.org/pdf/2206.15407. To find out more about the Shifts Project, please visit https://shifts.ai .

Readme

1. Data include raw data

2. Analysis using bootstrapping method to determine whether sleep mediate shift work and QOL

3. Baron and Kenny method to determine lifestyle factor mediate shift work and QOL

This dataset contains vegetation plot locations, descriptions, fractional cover, and sample identifier information from surveys conducted as part of the 2022 NASA Surface Biology Geology (SBG) High Frequency Time series (SHIFT) campaign. Surveys took place from 2022-02-23 to 2022-09-27 at the Jack and Laura Dangermond Preserve, Sedgwick Reserve, and Carpinteria Salt Marsh Reserve, which are located in Santa Barbara County, California, USA. This project collected field data contemporaneously with weekly flights of the NASA Airborne Visible-Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) facility instrument over the study areas. Plot information includes: plot tree subform, species lists, plot description, plot samples characterization, and plot location and contextual information. Related data packages contain additional biogeochemical, reflectance, and foliar data. Survey data and metadata are presented in comma-separated values (.csv) format along with survey plot polygons in GeoJSON (.geojson) format.

Global trade data of Shift under 87089980, 87089980 global trade data, trade data of Shift from 80+ Countries.

F1 and calibrated log likelihood. Results are averaged over 10 random seeds; standard deviation is given in the subscript. Tasks marked by * are subject to input data distribution shift while datasets marked by † are subject to annotator pool distribution shift. Methods marked by ‡ are those which estimate either worker skill or item difficulty. Aggregating the individual soft-labeling methods yields classifiers with consistently good uncertainty estimation (best on all text based tasks) and generally good raw performance in terms of F1 across tasks.

The data represents a shift in process variable presented in a control chart due to the sudden change in the machine setting.

This dataset contains photographs of the plots where field vegetation sampling was conducted during the 2022 NASA Surface Biology Geology (SBG) High Frequency Time series (SHIFT) campaign. Sampling occurred at the Jack and Laura Dangermond Preserve, Sedgwick Reserve, and Carpinteria Salt Marsh Reserve, which are located in Santa Barbara County, California, USA. Photographs were taken from 2022-02-23 to 2022-09-18. This project collected field data contemporaneously with weekly flights of Airborne Visible-Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) over the study areas. Related SHIFT data packages contain additional biogeochemical, reflectance, and foliar data.

Modern plankton high-throughput monitoring relies on deep learning classifiers for species recognition in water ecosystems. Despite satisfactory nominal performances, a significant challenge arises from the dataset shift, where performance drops during real-world deployment compared to ideal testing conditions. In our study, we integrate the ZooLake dataset, which consists of dark-field images of lake plankton, with manually-annotated images from 10 independent days of deployment, serving as test cells to benchmark out-of-dataset (OOD) performances. Our analysis reveals instances where classifiers, initially performing well in ideal conditions, encounter notable failures in real-world scenarios. For example, a MobileNet with a 92% nominal test accuracy shows a 77% OOD accuracy. We systematically investigate conditions leading to OOD performance drops and propose a preemptive assessment method to identify potential pitfalls when classifying new data, and pinpoint features in OOD images that adversely impact classification. We present a three-step pipeline: (i) identifying OOD degradation compared to nominal test performance, (ii) conducting a diagnostic analysis of degradation causes, and (iii) providing solutions. We find that ensembles of BEiT vision transformers, with targeted augmentations addressing OOD robustness, geometric ensembling, and rotation-based test-time augmentation, constitute the most robust model. It achieves an 83% OOD accuracy, with errors concentrated on container classes. Moreover, it exhibits lower sensitivity to dataset shift, and reproduces well the plankton abundances. Our proposed pipeline is applicable to generic plankton classifiers, contingent on the availability of suitable test cells. Implementation of this pipeline is anticipated to usher in a new era of robust classifiers, resilient to dataset shift, and capable of delivering reliable plankton abundance data. By identifying critical shortcomings and offering practical procedures to fortify models against dataset shift, our study contributes to the development of more reliable plankton classification technologies.

Traffic analytics, rankings, and competitive metrics for shift.com as of September 2025

We assess model performance using six datasets encompassing a broad taxonomic range. The number of species per dataset ranges from 28 to 239 (mean=118, median=94), and range shifts were observed over periods ranging from 20 to 100+ years. Each dataset was derived from previous evaluations of traits as range shift predictors and consists of a list of focal species, associated species-level traits, and a range shift metric.

This dataset holds full-resolution 3-band (true color) imagery acquired by NASA's Airborne Visible / Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) instrument. This imagery was collected as part of the Surface Biology and Geology High-Frequency Time Series (SHIFT) campaign which occurred during February to May, 2022, with a follow up activity for one week in September. The SHIFT campaign leveraged NASA's AVIRIS-NG facility instrument to collect VSWIR data at approximately a weekly cadence across a broad study area, enabling traceability analyses related to the science value of VSWIR revisits. AVIRIS-NG is a pushbroom spectral mapping system with high signal-to-noise ratio (SNR), designed and toleranced for high performance spectroscopy. AVIRIS-NG measures radiance at approximately 5-nm intervals in the Visible to Shortwave Infrared (VSWIR) spectral range from 380-2510 nm. The images in this dataset are true color (RGB) images from the wavelengths centered at approximately 808, 658, and 563 nm, subset from the full spectrum collected by AVIRIS-NG. The spatial resolution matches the native observed resolution (variable depending on the flightline, generally finer than 5 m and down to 2 m). There are two files for each flight line, one in PNG and one in georeferenced cloud-optimized GeoTIFF format; the GeoTIFF contains radiance floating point values while the PNG has been scaled and converted to integers.

This dataset contains the predicted prices of the asset shift over the next 16 years. This data is calculated initially using a default 5 percent annual growth rate, and after page load, it features a sliding scale component where the user can then further adjust the growth rate to their own positive or negative projections. The maximum positive adjustable growth rate is 100 percent, and the minimum adjustable growth rate is -100 percent.

Dataset

This dataset was created by Felix Fernandez

Contents