- title: 'DragonPaint: Rule Based Bootstrapping for Small Data with an Application to Cartoon Coloring' abstract: 'In this paper, we confront the problem of deep learning’s big labeled data requirements, offer a rule based strategy for extreme augmentation of small data sets and apply that strategy with the image to image translation model by \citetpix2pix:16 to automate cel style cartoon coloring with very limited training data. While our experimental results using geometric rules and transformations demonstrate the performance of our methods on an image translation task with industry applications in art, design and animation, we also propose the use of rules on partial data sets as a generalizable small data strategy, potentially applicable across data types and domains.' volume: 82 URL: https://proceedings.mlr.press/v82/greene18a.html PDF: http://proceedings.mlr.press/v82/greene18a/greene18a.pdf edit: https://github.com/mlresearch//v82/edit/gh-pages/_posts/2018-08-09-greene18a.md series: 'Proceedings of Machine Learning Research' container-title: 'Proceedings of The 4th International Conference on Predictive Applications and APIs' publisher: 'PMLR' author: - given: K. Gretchen family: Greene editor: - given: Claire family: Hardgrove - given: Louis family: Dorard - given: Keiran family: Thompson page: 1-9 id: greene18a issued: date-parts: - 2018 - 8 - 9 firstpage: 1 lastpage: 9 published: 2018-08-09 00:00:00 +0000 - title: 'Flexible and Scalable Deep Learning with MMLSpark' abstract: 'In this work we detail a novel open source library, called MMLSpark, that combines the flexible deep learning library Cognitive Toolkit, with the distributed computing framework Apache Spark. To achieve this, we have contributed Java Language bindings to the Cognitive Toolkit, and added several new components to the Spark ecosystem. In addition, we also integrate the popular image processing library OpenCV with Spark, and present a tool for the automated generation of PySpark wrappers from any SparkML estimator and use this tool to expose all work to the PySpark ecosystem. Finally, we provide a large library of tools for working and developing within the Spark ecosystem. We apply this work to the automated classification of Snow Leopards from camera trap images, and provide an end to end solution for the non-profit conservation organization, the Snow Leopard Trust.' volume: 82 URL: https://proceedings.mlr.press/v82/hamilton18a.html PDF: http://proceedings.mlr.press/v82/hamilton18a/hamilton18a.pdf edit: https://github.com/mlresearch//v82/edit/gh-pages/_posts/2018-08-09-hamilton18a.md series: 'Proceedings of Machine Learning Research' container-title: 'Proceedings of The 4th International Conference on Predictive Applications and APIs' publisher: 'PMLR' author: - given: Mark family: Hamilton - given: Sudarshan family: Raghunathan - given: Akshaya family: Annavajhala - given: Danil family: Kirsanov - given: Eduardo family: Leon - given: Eli family: Barzilay - given: Ilya family: Matiach - given: Joe family: Davison - given: Maureen family: Busch - given: Miruna family: Oprescu - given: Ratan family: Sur - given: Roope family: Astala - given: Tong family: Wen - given: ChangYoung family: Park editor: - given: Claire family: Hardgrove - given: Louis family: Dorard - given: Keiran family: Thompson page: 11-22 id: hamilton18a issued: date-parts: - 2018 - 8 - 9 firstpage: 11 lastpage: 22 published: 2018-08-09 00:00:00 +0000 - title: 'An Architecture and Domain Specific Language Framework for Repeated Domain-Specific Predictive Modeling' abstract: 'When repeatedly fitting related predictive models within the same domain, for similar problems, it’s helpful to have tools to support an efficient, high-quality workflow. This paper describes a theory of the architecture for such tools and for the interfaces among predictive models and other aspects of a software system. Additionally, it describes an open-source reference implementation of this design, written in R, focusing on a Domain Specific Language for one specific repeated predictive modeling task.' volume: 82 URL: https://proceedings.mlr.press/v82/harris18a.html PDF: http://proceedings.mlr.press/v82/harris18a/harris18a.pdf edit: https://github.com/mlresearch//v82/edit/gh-pages/_posts/2018-08-09-harris18a.md series: 'Proceedings of Machine Learning Research' container-title: 'Proceedings of The 4th International Conference on Predictive Applications and APIs' publisher: 'PMLR' author: - given: Harlan D. family: Harris editor: - given: Claire family: Hardgrove - given: Louis family: Dorard - given: Keiran family: Thompson page: 23-32 id: harris18a issued: date-parts: - 2018 - 8 - 9 firstpage: 23 lastpage: 32 published: 2018-08-09 00:00:00 +0000 - title: 'Marvin - Open source artificial intelligence platform' abstract: 'Marvin is an open source project that focuses on empowering data science teams to deliver industrial-grade applications supported by a high-scale, low-latency, language agnostic and standardized architecture platform, while simplifying the process of exploration and modeling. Building model-dependent applications in a robust way is not trivial, one is required to have knowledge in advanced areas of sciences like computing, statistics and math. Marvin aims at abstracting the complexities in the creation process of scalable, highly available, interoperable and maintainable predictive software.' volume: 82 URL: https://proceedings.mlr.press/v82/miguel18a.html PDF: http://proceedings.mlr.press/v82/miguel18a/miguel18a.pdf edit: https://github.com/mlresearch//v82/edit/gh-pages/_posts/2018-08-09-miguel18a.md series: 'Proceedings of Machine Learning Research' container-title: 'Proceedings of The 4th International Conference on Predictive Applications and APIs' publisher: 'PMLR' author: - given: Lucas B. family: Miguel - given: Daniel family: Takabayashi - given: Jose R. family: Pizani - given: Tiago family: Andrade - given: Brody family: West editor: - given: Claire family: Hardgrove - given: Louis family: Dorard - given: Keiran family: Thompson page: 33-44 id: miguel18a issued: date-parts: - 2018 - 8 - 9 firstpage: 33 lastpage: 44 published: 2018-08-09 00:00:00 +0000 - title: 'Prediction and Uncertainty Quantification of Daily Airport Flight Delays' abstract: 'One in four commercial airline flights is delayed, inconveniencing travelers and causing large financial losses for carriers. The ability to accurately predict delays would make travelers’ lives easier and save airlines money. In this work, we approach the problem of predicting flight delays using a Variational Long Short-Term Memory (LSTM) model. The model is trained to predict aggregate daily delays for U.S. airports using a combination of continuous and discrete variables, including weather, airport characteristics, and congestion. Monte Carlo Dropout, a Bayesian Deep Learning technique based on variational inference, is incorporated to provide planners with a well-calibrated prediction interval. We show that our Variational LSTM results in an average median absolute error of 5.8 minutes per day across 123 airports in the United States. Moreover, results show that predictive uncertainty is well explained through a calibration analysis.' volume: 82 URL: https://proceedings.mlr.press/v82/vandal18a.html PDF: http://proceedings.mlr.press/v82/vandal18a/vandal18a.pdf edit: https://github.com/mlresearch//v82/edit/gh-pages/_posts/2018-08-09-vandal18a.md series: 'Proceedings of Machine Learning Research' container-title: 'Proceedings of The 4th International Conference on Predictive Applications and APIs' publisher: 'PMLR' author: - given: Thomas family: Vandal - given: Max family: Livingston - given: Camen family: Piho - given: Sam family: Zimmerman editor: - given: Claire family: Hardgrove - given: Louis family: Dorard - given: Keiran family: Thompson page: 45-51 id: vandal18a issued: date-parts: - 2018 - 8 - 9 firstpage: 45 lastpage: 51 published: 2018-08-09 00:00:00 +0000