This paper identifies a problem with databases that support metadata. Previous research has proposed annotating values stored in a database with metadata, such as time, security, privacy, and quality. The metadata influences how a value is used. For example, sequenced temporal semantics proscribes comparing a value to one alive at a different time. But when values stored in a database are pulled into the realm of a programming language from a database through an API, web service, or in a user-defined function, a step-down transformation of the data occurs. The transformation strips the metadata, changing the semantics of the value. The metadata is discarded because a programming language process a scalar value, not one annotated with metadata. This metadata-related impedance mismatch between databases and programming languages limits the real-world adoption of databases that support metadata.

JSON keyword search searches the current versions of documents in a collection. However, JSON documents change over time due to edits. Some applications, such as data forensics and auditing, need to search past versions of documents and for changes to documents. This paper introduces a system called Temporal JSON Keyword Search (TJKS) for search in a collection of JSON documents that vary over time. TJKS lets users control which temporal slice, or part of the history, can be searched using a temporal search semantics; we support both of the major temporal semantics: sequenced and nonsequenced search. This paper presents the semantics of temporal JSON keyword search, discusses an efficient implementation, and evaluates the implementation. Our extensions are largely orthogonal to specific keyword search techniques, so this research provides a blueprint for extending keyword search to include time and potentially other kinds of metadata.

DBMS performance is dependent on many parameters, such as index selection, cache size, physical layout, and data partitioning. Some combinations of these parameters can lead to optimal performance for a given workload but selecting an optimal or near-optimal combination is challenging, especially for large databases with complex workloads. Among the hundreds of parameters, index selection is arguably the most critical parameter for performance. We propose a self-administered framework, called the Multiple Type and Attribute Index Selector (MANTIS), that automatically selects near-optimal indexes. The framework advances the state-of-the-art index selection by considering both multi-attribute and multiple types of indexes within a bounded storage size constraint, a combination not previously addressed. MANTIS combines supervised and reinforcement learning, a Deep Neural Network recommends the type of index for a given workload while a Deep Q-Learning network recommends the multi-attribute aspect. MANTIS is sensitive to storage cost constraints and incorporates noisy rewards in its reward function for better performance. Our experimental evaluation shows that MANTIS outperforms the current state-of-art methods by an average of 9.53% [email protected]

A whole farm carbon model estimates the emissions of greenhouse gasses (GHGs) based on information for a farm. We analyzed two models, Holos whole-farm and COMET-Farm, by running the models on random inputs and building predictive models from the runs. Holos estimates GHG emissions for a particular year based on crop and animal agriculture input, while COMET-farm adds past and future farm management practices. Users of the models must manually enter farm data through a graphical user interface (GUI), which is a good method for a single farm, but makes it infeasible to calculate GHG emissions over hundreds to thousands of farms. So we automated the interface and generated random farm scenarios within ranges given by experts. We scraped the estimated carbon footprint from thousands of runs of the models and used algorithms to build predictive models that have high accuracy. By reverse engineering the whole-farm carbon models we were able to determine which farm management practices in each whole farm carbon model have the biggest impact on GHG emissions. This can help farmers and rural planners change farm management practices to decrease GHG emissions.

Coronavirus Disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 virus (SARS-CoV-2). The virus transmits rapidly; it has a basic reproductive number (R0) of 2.2-2.7. In March 2020, the World Health Organization declared the COVID-19 outbreak a pandemic. COVID-19 is currently affecting more than 200 countries with 6M active cases. An effective testing strategy for COVID-19 is crucial to controlling the outbreak but the demand for testing surpasses the availability of test kits that use Reverse Transcription Polymerase Chain Reaction (RT-PCR). In this paper, we present a technique to screen for COVID-19 using artificial intelligence. Our technique takes only seconds to screen for the presence of the virus in a patient. We collected a dataset of chest X-ray images and trained several popular deep convolution neural network-based models (VGG, MobileNet, Xception, DenseNet, InceptionResNet) to classify the chest X-rays. Unsatisfied with these models, we then designed and built a Residual Attention Network that was able to screen COVID-19 with a testing accuracy of 98% and a validation accuracy of 100%. A feature maps visual of our model show areas in a chest X-ray which are important for classification. Our work can help to increase the adaptation of AI-assisted applications in clinical practice. The code and dataset used in this project are available at https://github.com/vishalshar/covid-19-screening-using-RAN-on-X-ray-images.

Online crowdfunding platforms have given creators new opportunities to obtain funding. Despite the popularity and success of many projects on the platforms, the quality of crowd-funded products in the market (e.g., Amazon) was not statistically and scientifically evaluated yet. To fill the gap, in this paper, we (i) compare crowdfunded products with traditional products in terms of their ratings in the largest e-commerce market, Amazon; (ii) analyze characteristics of the successful products (received ≥4 star) and unsuccessful products (received <; 4 star); and (iii) build machine learning models in three different stages, which predict whether a crowdfunded product will receive high star ratings or not. Our experimental results show that crowdfunded products, on average, received lower rating than traditional products. Our predictive models effectively identify which product will receive high star-ratings from customers on Amazon. The datasets used in this paper will be available at http://web.cs.wpi.edu/~kmlee/data.html.

Planning a trip can be a tedious task. One has to search for what places to visit at a destination (i.e. area) and what time to visit the destination. Sometimes this can be a time-consuming task because there are too much information available, and it is hard for one to choose which information to t