Why GBSD Should Consider Machine Learning (ML) and Causal Learning (CL)
Abstract
BLUF. 1. GBSD is definitely a software system. 2. As such, many large streams of sensor and other data will be available. 3. This software system offers many benefits but also faces critical challenges and threats. 4. Machine learning will digest and model these large streams of data, if GBSD captures.and stores the data for later modeling. 5. Causal learning can actually determine cause-effect relationships in data as opposed.to spurious correlation, there by offering results not possible via traditional statistics and machine learning. 6. SEI is poised to contribute to the GBSD Program Office in specifying the data capture and storage needed, and the adoption of these new technologies with oversight and leadership with the contractors. 7. Almost all ilities represent low-hanging fruit opportunities for ML and CL. What is Machine Learning? Basically a more sophisticated form of correlation, association and pattern recognition. Can accommodate and needs Big Data, e.g. large volumes of streams of data. Forms include: 1. Unsupervised machine learning, e.g. to explore relationships between factors. 2. Supervised machine learning, e.g. to predict outcome(s). 3. Deep Learning (DL), e.g. a layered network to better identify and learn patterns. 4. Reinforcement Learning (RL), e.g. a network that helps learn actions to maximize a reward, sort of an optimization approach. 5. Generative Adversarial Networks (GANs), e.g. a set of networks that can interact with each other to generate additional data based on what each network learns from the other.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2019
- Accession Number
- AD1088914
Entities
People
- Robert W. Stoddard
Organizations
- Carnegie Mellon University