Challenges in Machine Generation of Analytic Products from Multi-Source Data

The Intelligence Community Studies Board of the National Academies of Sciences, Engineering, and Medicine convened a workshop on August 9-10, 2017 to examine challenges in machine generation of analytic products from multi-source data. Workshop speakers and participants discussed research challenges related to machine-based methods for generating analytic products and for automating the evaluation of these products, with special attention to learning from small data, using multi-source data, adversarial learning, and understanding the human-machine relationship. This publication summarizes the presentations and discussions from the workshop.Table of ContentsFront Matter1 Introduction2 Session 1: Plenary3 Session 2: Machine Learning from Image, Video, and Map Data4 Session 3: Machine Learning from Natural Languages5 Session 4: Learning from Multi-Source Data6 Session 5: Learning from Noisy, Adversarial Inputs7 Session 6: Learning from Social Media8 Session 7: Humans and Machines Working Together with Big Data9 Session 8: Use of Machine Learning for Privacy Ethics10 Session 9: Evaluation of Machine-Generated Products11 Session 10: Capability Technology MatrixAppendixesAppendix A: Biographical Sketches of Workshop Planning CommitteeAppendix B: Workshop AgendaAppendix C: Workshop Statement of TaskAppendix D: Capability Technology TablesAppendix E: Acronyms