Application Programming Interface Research Articles

New Sustainable Fintech Business Models Created by Open Application Programming Interface Technology: A Case Study of Korea’s Open Banking Application Programming Interface Platform

Fintech facilitates financial inclusion by introducing new sustainable and accessible business models in developing countries. Open banking API technology is used by various fintech businesses in both developing and developed countries, with varying effects on financial markets. It changes the financial market distribution structure by separating financial product manufacturing and distribution and intensifying competition between traditional financial institutions and fintech companies. Fintech companies innovate by using this tool to create new business models. In December 2019, Korea established a standardized “open banking API platform”, sharing an interbank payment network with fintech companies. It has since shown explosive growth, surpassing the trading volume of the UK, the country that introduced open banking APIs for the first time. This study analyzes new business models fintech companies created using this API technology. Based on existing literature, statistics from the platform operator (the Korea Financial Telecommunications and Clearings Institute), and investigations of 30 Korean fintech mobile applications, this study analyzes new fintech business models that provide financial services using this platform. These companies create sustainable business models by combining multiple APIs. Four representative business models (simple funds transfer, simple payment, cross-border remittance, and asset management) are analyzed to reveal how fintech companies create business models with open APIs.

ChatGPT versus a customized AI chatbot (Anatbuddy) for anatomy education: A comparative pilot study.

Large Language Models (LLMs) have the potential to improve education by personalizing learning. However, ChatGPT-generated content has been criticized for sometimes producing false, biased, and/or hallucinatory information. To evaluate AI's ability to return clear and accurate anatomy information, this study generated a custom interactive and intelligent chatbot (Anatbuddy) through an Open AI Application Programming Interface (API) that enables seamless AI-driven interactions within a secured cloud infrastructure. Anatbuddy was programmed through a Retrieval Augmented Generation (RAG) method to provide context-aware responses to user queries based on a predetermined knowledge base. To compare their outputs, various queries (i.e., prompts) on thoracic anatomy (n = 18) were fed into Anatbuddy and ChatGPT 3.5. A panel comprising three experienced anatomists evaluated both tools' responses for factual accuracy, relevance, completeness, coherence, and fluency on a 5-point Likert scale. These ratings were reviewed by a third party blinded to the study, who revised and finalized scores as needed. Anatbuddy's factual accuracy (mean ± SD = 4.78/5.00 ± 0.43; median = 5.00) was rated significantly higher (U = 84, p = 0.01) than ChatGPT's accuracy (4.11 ± 0.83; median = 4.00). No statistically significant differences were detected between the chatbots for the other variables. Given ChatGPT's current content knowledge limitations, we strongly recommend the anatomy profession develop a custom AI chatbot for anatomy education utilizing a carefully curated knowledge base to ensure accuracy. Further research is needed to determine students' acceptance of custom chatbots for anatomy education and their influence on learning experiences and outcomes.

Quantification of Automatic Speech Recognition System Performance on d/Deaf and Hard of Hearing Speech.

To evaluate the performance of commercial automatic speech recognition (ASR) systems on d/Deaf and hard-of-hearing (d/Dhh) speech. A corpus containing 850 audio files of d/Dhh and normal hearing (NH) speech from the University of Memphis Speech Perception Assessment Laboratory was tested on four speech-to-text application program interfaces (APIs): Amazon Web Services, Microsoft Azure, Google Chirp, and OpenAI Whisper. We quantified the Word Error Rate (WER) of API transcriptions for 24 d/Dhh and nine NH participants and performed subgroup analysis by speech intelligibility classification (SIC), hearing loss (HL) onset, and primary communication mode. Mean WER averaged across APIs was 10 times higher for the d/Dhh group (52.6%) than the NH group (5.0%). APIs performed significantly worse for "low" and "medium" SIC (85.9% and 46.6% WER, respectively) as compared to "high" SIC group (9.5% WER, comparable to NH group). APIs performed significantly worse for speakers with prelingual HL relative to postlingual HL (80.5% and 37.1% WER, respectively). APIs performed significantly worse for speakers primarily communicating with sign language (70.2% WER) relative to speakers with both oral and sign language communication (51.5%) or oral communication only (19.7%). Commercial ASR systems underperform for d/Dhh individuals, especially those with "low" and "medium" SIC, prelingual onset of HL, and sign language as primary communication mode. This contrasts with Big Tech companies' promises of accessibility, indicating the need for ASR systems ethically trained on heterogeneous d/Dhh speech data. 3 Laryngoscope, 2024.

Large Language Models in Otolaryngology Residency Admissions: A Random Sampling Analysis.

To investigate potential demographic bias in artificial intelligence (AI)-based simulations of otolaryngology, residency selection committee (RSC) members tasked with selecting one applicant among candidates with varied racial, gender, and sexual orientations. This study employed random sampling of simulated RSC member decisions using a novel Application Programming Interface (API) to virtually connect to OpenAI's Generative Pre-Trained Transformers (GPT-4 and GPT-4o). Simulated RSC members with diverse demographics were tasked with ranking to match 1 applicant among 10 with varied racial, gender, and sexual orientations. All applicants had identical qualifications; only demographics of the applicants and RSC members were varied for each simulation. Each RSC simulation ran 1000 times. Chi-square tests analyzed differences across categorical variables. GPT-4o simulations additionally requested a rationale for each decision. Simulated RSCs consistently showed racial, gender, and sexual orientation bias. Most applicant pairwise comparisons showed statistical significance (p < 0.05). White and Black RSCs exhibited greatest preference for applicants sharing their own demographic characteristics, favoring White and Black female applicants, respectively, over others (all pairwise p < 0.001). Asian male applicants consistently received lowest selection rates. Male RSCs favored White male and female applicants, while female RSCs preferred LGBTQIA+, White and Black female applicants (all p < 0.05). High socioeconomic status (SES) RSCs favored White female and LGBTQIA+ applicants, while low SES RSCs favored Black female and LGBTQIA+ applicants over others (all p < 0.001). Results from the newest iteration of the LLM, ChatGPT-4o, indicated evolved selection preferences favoring Black female and LGBTQIA+ applicants across all RSCs, with the rationale of prioritizing inclusivity given in >95% of such decisions. Utilizing publicly available LLMs to aid in otolaryngology residency selection may introduce significant racial, gender, and sexual orientation bias. Potential for significant and evolving LLM bias should be appreciated and minimized to promote a diverse and representative field of future otolaryngologists in alignment with current workforce data. N/A Laryngoscope, 2024.

Reusable rate of wooden structural members and its optimization determined by application programming interface (API)

ABSTRACT Carbon is stored within the structure of wood; however, burnt or landfilled wooden components releases carbon back into the atmosphere. Increasing the reusability rate of wooden components potentially prolongs carbon storage in wood. This study investigated the reusability rate of wooden brace members from a high-rise wooden building by using an application programming interface (API) integrated with Revit. This study demonstrated a method for improving the reusability rate by optimizing the brace cross-sections. The API assists designers in determining the reusability rate early in the design process by computing and representing reusability rates based on the designated converted members. Optimization enhanced the reusability of specific sections, with a reusability rate of approximately 80% achieved for the optimized cross-section area of 28.0 × 51.6 cm2 compared with 62% for the benchmark cross-section area of 38 × 38 cm2. Additionally, a preliminary structural assessment of the optimized wooden brace sections was conducted, which confirmed that the optimization did not affect the structural integrity. Consequently, selecting optimized wooden brace sections with high reusability rates by using the API extends the service life of wood and prolongs carbon sequestration, thereby mitigating carbon dioxide emission into the atmosphere.

Human BioMolecular Atlas Program (HuBMAP): 3D Human Reference Atlas Construction and Usage.

The Human BioMolecular Atlas Program (HuBMAP) aims to construct a reference 3D structural, cellular, and molecular atlas of the healthy adult human body. The HuBMAP Data Portal (https://portal.hubmapconsortium.org) serves experimental datasets and supports data processing, search, filtering, and visualization. The Human Reference Atlas (HRA) Portal (https://humanatlas.io) provides open access to atlas data, code, procedures, and instructional materials. Experts from more than 20 consortia are collaborating to construct the HRA's Common Coordinate Framework (CCF), knowledge graphs, and tools that describe the multiscale structure of the human body (from organs and tissues down to cells, genes, and biomarkers) and to use the HRA to understand changes that occur at each of these levels with aging, disease, and other perturbations. The 6th release of the HRA v2.0 covers 36 organs with 4,499 unique anatomical structures, 1,195 cell types, and 2,089 biomarkers (e.g., genes, proteins, lipids) linked to ontologies and 2D/3D reference objects. New experimental data can be mapped into the HRA using (1) three cell type annotation tools (e.g., Azimuth) or (2) validated antibody panels (OMAPs), or (3) by registering tissue data spatially. This paper describes the HRA user stories, terminology, data formats, ontology validation, unified analysis workflows, user interfaces, instructional materials, application programming interface (APIs), flexible hybrid cloud infrastructure, and previews atlas usage applications.

Adolescents' unhealthy snacking behaviour during the school journey and the association with transport modes and food outlets along the school route.

Active transport to and from school provides an opportunity for adolescents to engage in physical activity, but travelling through an obesogenic environment may have unintended consequences on their snacking behaviour. This study aimed to: (1) identify whether adolescents' unhealthy snacking behaviour was associated with school transport modes and food outlets on their estimated school route and (2) explore whether food outlet density on the school route differed between school transport modes. Adolescents (n = 406; aged 15.1 ± 1.4 years; 50.7% boys; 63.5% New Zealand European) from all 12 secondary schools in Dunedin city, Aotearoa New Zealand, completed an online survey. School transport modes (active, motorised or mixed) and unhealthy snacking data were collected. Food outlet data were collected using Google Places Application Programming Interface (API). Home-to-school route and distance were estimated using geographical information system (GIS) analysis based on a walkable road network. Data were analysed using the chi-square test and logistic regression. Overall, 26.4% of adolescents reported purchasing and consuming unhealthy snacks or soft drinks on the way to school and 41.4% from school. The odds of unhealthy snacking during the school journey was higher among mixed transport users than active transport users on the way to (odds ratio (OR) = 2.73, 95% confidence interval (CI) = 1.39-5.36) and from school (OR = 2.68, 95% CI = 1.40-5.13). No differences were observed by food outlet type. There were no food outlets on the estimated school route for 44.8% of adolescents. The presence of more than one food outlet per kilometre of the estimated school route differed between active (38.7%), motorised (42.6%) and mixed transport users (46.4%; p = 0.006). School transport modes were significantly associated with adolescents' unhealthy snacking behaviour and food outlets on their school journey. Policy measures which minimise exposure to unhealthy food outlets may reduce unhealthy snacking among adolescents.

An Introduction to the Global Registry of Scientific Collections (GRSciColl)

The Global Registry of Scientific Collections (GRSciColl) is a comprehensive community-curated repository of scientific collections, their host institutions, and associated staff members. GRSciColl was adopted by the Global Biodiversity Information Facility (GBIF) in 2018, after being developed and hosted at the Smithsonian Institution. GRSciColl houses data on various institutional or personal scientific collections such as natural history collections, archeological collections, and cell culture collections. Users can access information about the collection location, content, accessibility status, coverage (e.g., geographic and taxonomic), and contacts. Additionally, it serves as a tool for enhancing data interoperability by including institution and collection codes and identifiers, as well as external identifiers like Wikidata Q numbers whenever feasible. The digitized specimens published on GBIF.org are linked with GRSciColl institutions and collections based on these codes and identifiers. These specimens are then aggregated to generate metrics and dashboards, which display customized multiple tables and visualization, regardless of how they were published on GBIF. Most of the data in GRSciColl is sourced from Index Herbariorum, the Integrated Digitized Biocollections (iDigBio), and national registries. The information can also be improved by direct contributions from a community of editors or the suggestions made by the users. As of 19 April 2024, GRSciColl provided information on 7,678 scientific collections and 7,906 institutions. However, GRSciColl includes numerous legacy entries, with many institutions lacking associated collections. This discrepancy arises partly from misregistrations, where institution entries should be collections. Additionally, some national registries or surveys provided institution-level information that did not align with the GRSciColl schema for collections. Furthermore, some records pertain to historical institutions that no longer maintain collections. The most up-to-date entries are discoverable and accessible via the website: https://scientific-collections.gbif.org and the GRSciColl Application Programming Interface (API). In this presentation, we will show you how to navigate the GRSciColl website and API to help you find relevant institutions, collections, and specimen records and look up codes and identifiers. We will cover any new functionality developed in the context of the GRSciColl 2023-2024 road map, which aims to support richer, structured descriptions as well as the ability to upload an inventory of species.

GPT is an effective tool for multilingual psychological text analysis

The social and behavioral sciences have been increasingly using automated text analysis to measure psychological constructs in text. We explore whether GPT, the large-language model (LLM) underlying the AI chatbot ChatGPT, can be used as a tool for automated psychological text analysis in several languages. Across 15 datasets (n = 47,925 manually annotated tweets and news headlines), we tested whether different versions of GPT (3.5 Turbo, 4, and 4 Turbo) can accurately detect psychological constructs (sentiment, discrete emotions, offensiveness, and moral foundations) across 12 languages. We found that GPT (r = 0.59 to 0.77) performed much better than English-language dictionary analysis (r = 0.20 to 0.30) at detecting psychological constructs as judged by manual annotators. GPT performed nearly as well as, and sometimes better than, several top-performing fine-tuned machine learning models. Moreover, GPT's performance improved across successive versions of the model, particularly for lesser-spoken languages, and became less expensive. Overall, GPT may be superior to many existing methods of automated text analysis, since it achieves relatively high accuracy across many languages, requires no training data, and is easy to use with simple prompts (e.g., "is this text negative?") and little coding experience. We provide sample code and a video tutorial for analyzing text with the GPT application programming interface. We argue that GPT and other LLMs help democratize automated text analysis by making advanced natural language processing capabilities more accessible, and may help facilitate more cross-linguistic research with understudied languages.

Early Ransomware Detection with Deep Learning Models

Ransomware is a growing-in-popularity type of malware that restricts access to the victim’s system or data until a ransom is paid. Traditional detection methods rely on analyzing the malware’s content, but these methods are ineffective against unknown or zero-day malware. Therefore, zero-day malware detection typically involves observing the malware’s behavior, specifically the sequence of application programming interface (API) calls it makes, such as reading and writing files or enumerating directories. While previous studies have used machine learning (ML) techniques to classify API call sequences, they have only considered the API call name. This paper systematically compares various subsets of API call features, different ML techniques, and context-window sizes to identify the optimal ransomware classifier. Our findings indicate that a context-window size of 7 is ideal, and the most effective ML techniques are CNN and LSTM. Additionally, augmenting the API call name with the operation result significantly enhances the classifier’s precision. Performance analysis suggests that this classifier can be effectively applied in real-time scenarios.

Negative sentiments toward anterior cruciate ligament injury prevention outweigh positive awareness discussions on social media

ObjectivesThe aim of this study is to understand the current public discussion surrounding anterior cruciate ligament (ACL) injury prevention on social media and determine factors that influence levels of public engagement. MethodsWe performed a qualitative and quantitative cross-sectional analysis of ACL injury prevention techniques discussed on social media via the Twitter application programming interface (API). The Twitter API was queried from inception to May 2023 using keywords related to ACL injury and prevention. We conducted a thematic analysis of the posts and performed a sentiment analysis using natural language processing. A multivariable regression model was used to identify metadata that predicted higher engagement (media, links, tagging, hashtags). ResultsA subset of 1823 unique posts was analyzed from 1701 unique accounts. Most posts were raising awareness about ACL injury prevention (n ​= ​733, 40.2%), followed by opinions on the topic (n ​= ​390, 21.4%), specific prevention techniques (n ​= ​289, 15.9%), personal experiences (n ​= ​272, 14.9%), and research (n ​= ​139, 7.6%). The majority consisted of posts from patients or caregivers (n ​= ​948, 55.7%), whereas healthcare providers accounted for 14.7% of posts. Posts containing media increased Tweet engagement count by an average of 6.1 (95% CI 2.8 to 9.4, p ​= ​0.00033) and posts discussing personal opinions increased engagement by 6.7 (95% CI 3.5 to 9.8, p ​= ​0.00004). On sentiment analysis of all included Tweets, 822 (45.1%) posts were positive, 309 (17.0%) were negative, and 692 (38.0%) were neutral. Sentiments expressed in posts related to ACL prevention were 2.8 times more negative compared to those discussing raising awareness. ConclusionsThere is active discussion about ACL injury prevention on Twitter. The use of visual media increased public engagement. We identified a potential knowledge gap between the available prevention techniques and the perspectives of athletes, highlighting the need for healthcare professionals to enhance their engagement with ACL injury prevention on social media. Level of evidenceIII.

Design and architecture of the CARA infrastructure for visualising and benchmarking patient data from general practice

ObjectiveCollaborate, Analyse, Research and Audit (CARA) project set out to provide an infrastructure to enable Irish general practitioners (GPs) to use their routinely collected patient management software (PMS) data to...

DiSSCover the Potential of FAIR Digital Object Annotations and How You Can Use Them!

The infrastructure for the Distributed System of Scientific Collections (DiSSCo) is in full development. Work within the DiSSCo Transition Project has been focused on building infrastructure, creating data models, and setting up Application Programming Interfaces (APIs) (Koureas et al. 2024). In the past years, DiSSCo has presented this work at different Biodiversity Information Standards (TDWG) conferences (Leeflang and Addink 2023, Leeflang et al. 2022, Addink et al. 2021). In this year’s session, we would like to focus on the human-facing application: DiSSCover. DiSSCover is the graphical user interface through which users can interact with Findable, Accessible, Interoperable and Reusable (FAIR) Digital Objects (FDOs), facilitating the curation and enhancement of specimen data (Islam 2024). Development started in 2022 and is ongoing. The interface acts as a gateway into the DiSSCo infrastructure, providing access to digital specimens and media. Extracted from the core DiSSCo API, the data is converted into an easily readable format and made discoverable through a diverse set of filters. DiSSCover’s main focus is to allow users to make annotations upon the data. Through the concept of annotations, we connect expert and machine-generated information, to create extended digital specimens (Hardisty et al. 2022), e.g, by creating linkages to other infrastructures, correcting or adding new information, or by triggering machine annotation services. Machine annotation services are automated scalable tools that run in the background and automatically curate and extend the specimen (Addink et al. 2023). Human users will remain important, as all annotations made by machine annotation services can be reviewed by a trusted person. Annotations are Fair Digital Objects and target a specific part of a specimen, be it a data fragment or an associated media file. At the heart of DiSSCover lies the Open Digital Specimen data specification (Leeflang and Addink 2023). It tries to harmonise multiple data standards into one generic specification based on the new Global Biodiversity Information Facility (GBIF) Unified Model (Robertson et al. 2022). The data is stored as JavaScript Object Notation (JSON) based on JSON Schemas (Anonymous 2024). Annotations are linked to specific data attributes using a JSON-path as the identifier. Data attributes can be individual terms, collections of terms called classes, or the whole object. This creates a flexible but complex data structure, the basis for which we used the World Wide Web Consortium (W3C) web annotation data model (Sanderson et al. 2017). The W3C Web annotation data model contains two main components: the target and body. The target specifies which data attribute the annotation is made on, for example, the term: ‘ods:specimenName’. This is a local term within the open Digital Specimen namespace (ods), which holds the accepted name of the digital specimen. The annotation body holds the value(s) that are appended to the digital specimen, and differ based upon the annotation motivation. DiSSCo recognises five different annotation motivations: addition, modification, comment, assessment and deletion, each of which has its own unique function. This creates a flexible structure that should be able to handle any information the user wants to add to the object. The challenge of DiSSCover is to preserve the complex structure of annotations, whilst making it convenient for users to work with. The session will provide a look at the different kinds of annotations and their use from a practical perspective. A demonstration of DiSSCover will show how users can create annotations, providing knowledge about the process that will give shape to DiSSCo’s main goal of enriching natural history data.

Transforming Wikipedia text into audio through machine learning

Purpose The purpose of this study is to navigate the process of transforming Wikipedia articles into audio files for library readers. The system provides a feasible manner of listening to Wikipedia content, accommodating diverse learning preferences and enlarging knowledge in education society. Design/methodology/approach This framework has been constructed using the Python programming languages in the Linux operating platform. Application programming interface and Google text-to-speech (TTS) are required as additional software packages to design this prototype system. Transform any Wikipedia pages into audio files through Wikitrola for libraries and information centers. Wikipedia articles are directly transformed into audio, as these integrate the content seamlessly for the user experience. The whole system has been designed and configured on the basis of machine learning to provide dynamic services among the readers. Findings The viewer could use the machine learning system to turn Wikipedia articles into audio files, allowing them to listen to Wikipedia content in audio format. This would make information more accessible and adaptable to diverse learning modes, allowing written content to be engaged in novel and visionary ways. Originality/value The insightful observation in connection with the paper is that it shows how to convert text-based material into audio through the Google TTS and machine learning Python programming and finally incorporate them in Wikipedia articles. A harmonious system of information dissemination and technical education is established. This approach shows the effectiveness of imagination and the use of programming tools to enhance learning and knowledge-seeking processes.

User Privacy Protection via Windows Registry Hooking and Runtime Encryption.

The Windows registry contains a plethora of information in a hierarchical database. It includes system-wide settings, user preferences, installed programs, and recently accessed files and maintains timestamps that can be used to construct a detailed timeline of user activities. However, these data are unencrypted and thus vulnerable to exploitation by malicious actors who gain access to this repository. To address this security and privacy concern, we propose a novel approach that efficiently encrypts and decrypts sensitive registry data in real time. Our developed proof-of-concept program intercepts interactions between the registry's application programming interfaces (APIs) and other Windows applications using an advanced hooking technique. This enables the proposed system to be transparent to users without requiring any changes to the operating system or installed software. Our approach also implements the data protection API (DPAPI) developed by Microsoft to securely manage each user's encryption key. Ultimately, our research provides an enhanced security and privacy framework for the Windows registry, effectively fortifying the registry against security and privacy threats while maintaining its accessibility to legitimate users and applications.

Development of Remote Condition Monitoring System for Panauti Hydropower Station

Abstract Condition monitoring is the process of predicting faults in machine through the measurement and analysis of pre-identified parameters. This paper aims to present the features and process of development of a real time condition monitoring system installed at Panauti Hydropower station. In total, seven different sensors including Pressure, Temperature, Guide Vane Position and Tri-axial Vibration sensor were installed at the power plant. The locations of the pressure sensors are at the inlet of the turbine and the outlet of the draft tube. Tri-axial vibration sensor was installed at the turbine shaft to capture the turbine vibrations at different operating conditions. The temperature sensors provide the temperature data from the bearings and housing of a generator. The sensors were connected to the NI-DAQ (data acquisition system) hardware including current measuring module and sound and vibration module integrated with the LabVIEW program. The data acquisition system is capable of recording the pressure pulsations at the outlet of draft tube. The monitoring system processes the data from sensors and uploads the relevant information to the Kathmandu University’s central server. A remote monitoring program in LabVIEW has been developed which uses an Application Programming Interface (API) to retrieve the information from the central server in the real time, enabling the monitoring of the power plant from any location.

Single/Multi-Source Black-Box Domain Adaption for Sensor Time Series Data.

Unsupervised domain adaption (UDA), which transfers knowledge from a labeled source domain to an unlabeled target domain, has attracted tremendous attention in many machine learning applications. Recently, there have been attempts to apply domain adaption for sensor time series data, such as human activity recognition and gesture recognition. However, existing methods suffer from some drawbacks that hinder further performance improvement. They often require access to source data or source models during training, which is unavailable in some fields because of privacy protection and storage limit. Typically, the source domains may only provide an application programming interface (API) for the target domain to call. On the other hand, current UDA methods have not considered the temporal consistency and low-signal-to-noise ratio (SNR) of sensor time series. To address the challenges, this article presents a black-box domain adaption framework for sensor time series data (B2TSDA). First, we propose a single/multi-source teacher-student learning framework to distill the knowledge from the source domains to a customized target model. Then we design a new temporal consistency loss by combining an adaptive mask method and dynamic threshold method to maintain consistent temporal information and balance the learning difficulties of different classes. For the multisource black-box domain adaption, we further propose a Shapley-enhanced method to determine the contribution of each source domain. Experimental results on both single-source and multisource domain adaption show that our framework has superior performance compared to other black-box UDA methods.

TakeMeHome: A Novel Method for Reaching Previously Untested People Through Online Ordering and Self-Collect HIV and STI Testing.

Despite national testing guidelines, rates of testing for HIV, sexually transmitted infections (STIs), and hepatitis C remain lower than recommended for men who have sex with men in the United States. To help address this, the TakeMeHome (TMH) program was started in March 2020 by Building Healthy Online Communities, a consortium of public health organizations and dating apps, to work with health departments to increase access to HIV testing for men who have sex with men on dating apps. Users of participating dating apps were sent messages about opportunities for testing with self-collected specimens through TMH. Program users were eligible to receive test kits if they lived in a participating zip code and were aged at least 18 years. Users who were interested in testing could order kits to be mailed to them for laboratory-based testing of HIV, hepatitis C, chlamydia, gonorrhea, and/or syphilis, depending on risk and availability in their zip code. Orders were sent via application programming interface to Molecular Testing Labs for fulfillment; kits were provided at no cost to the program user. Within approximately 24 hours of order receipt, Molecular Testing Labs mailed program users a kit with required collection supplies, directions, and a link to a video instruction for self-collection. Program users received an automated email after testing was complete with a link to access results through their online account. Individuals with positive results on any of the relevant tests were directed to additional information and to local providers for linkage to additional testing or treatment, depending on local protocols. The positivity rate of specimens processed through TMH was 1.4% for HIV, 0.6% for hepatitis C, and 2.9% for all STIs combined. The per-person positivity rate was 15.3% across all STIs. The TMH program demonstrates that self-collected laboratory-processed testing is feasible and effective at identifying new HIV and STI cases.

MotorNet, a Python toolbox for controlling differentiable biomechanical effectors with artificial neural networks.

Artificial neural networks (ANNs) are a powerful class of computational models for unravelling neural mechanisms of brain function. However, for neural control of movement, they currently must be integrated with software simulating biomechanical effectors, leading to limiting impracticalities: (1) researchers must rely on two different platforms and (2) biomechanical effectors are not generally differentiable, constraining researchers to reinforcement learning algorithms despite the existence and potential biological relevance of faster training methods. To address these limitations, we developed MotorNet, an open-source Python toolbox for creating arbitrarily complex, differentiable, and biomechanically realistic effectors that can be trained on user-defined motor tasks using ANNs. MotorNet is designed to meet several goals: ease of installation, ease of use, a high-level user-friendly application programming interface, and a modular architecture to allow for flexibility in model building. MotorNet requires no dependencies outside Python, making it easy to get started with. For instance, it allows training ANNs on typically used motor control models such as a two joint, six muscle, planar arm within minutes on a typical desktop computer. MotorNet is built on PyTorch and therefore can implement any network architecture that is possible using the PyTorch framework. Consequently, it will immediately benefit from advances in artificial intelligence through PyTorch updates. Finally, it is open source, enabling users to create and share their own improvements, such as new effector and network architectures or custom task designs. MotorNet's focus on higher-order model and task design will alleviate overhead cost to initiate computational projects for new researchers by providing a standalone, ready-to-go framework, and speed up efforts of established computational teams by enabling a focus on concepts and ideas over implementation.

MotorNet, a Python toolbox for controlling differentiable biomechanical effectors with artificial neural networks

Artificial neural networks (ANNs) are a powerful class of computational models for unravelling neural mechanisms of brain function. However, for neural control of movement, they currently must be integrated with software simulating biomechanical effectors, leading to limiting impracticalities: (1) researchers must rely on two different platforms and (2) biomechanical effectors are not generally differentiable, constraining researchers to reinforcement learning algorithms despite the existence and potential biological relevance of faster training methods. To address these limitations, we developed MotorNet, an open-source Python toolbox for creating arbitrarily complex, differentiable, and biomechanically realistic effectors that can be trained on user-defined motor tasks using ANNs. MotorNet is designed to meet several goals: ease of installation, ease of use, a high-level user-friendly application programming interface, and a modular architecture to allow for flexibility in model building. MotorNet requires no dependencies outside Python, making it easy to get started with. For instance, it allows training ANNs on typically used motor control models such as a two joint, six muscle, planar arm within minutes on a typical desktop computer. MotorNet is built on PyTorch and therefore can implement any network architecture that is possible using the PyTorch framework. Consequently, it will immediately benefit from advances in artificial intelligence through PyTorch updates. Finally, it is open source, enabling users to create and share their own improvements, such as new effector and network architectures or custom task designs. MotorNet’s focus on higher-order model and task design will alleviate overhead cost to initiate computational projects for new researchers by providing a standalone, ready-to-go framework, and speed up efforts of established computational teams by enabling a focus on concepts and ideas over implementation.