Knowledge-Driven Machine Learning
Jayaraman Jayaraman Thiagarajan | 21-ERD-012
Executive Summary
We will develop a novel knowledge-driven learning framework for machine learning that can leverage information from knowledge graphs without explicit supervision. The resulting advanced techniques can lead to unprecedented opportunities to improve the effectiveness of Lawrence Livermore National Laboratory’s work in several strategic mission areas including biomedicine and nuclear threat reduction.
Publications, Presentations, and Patents
Ramadan, Tarek, Tanzima Z. Islam, Chase Phelps, Nathan Pinnow, and Jayaraman J. Thiagarajan. "Comparative Code Structure Analysis using Deep Learning for Performance Prediction." In 2021 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), pp. 151-161. 2021.
Shanthamallu, Uday Shankar, Jayaraman J. Thiagarajan, and Andreas Spanias. "Uncertainty-Matching Graph Neural Networks to Defend Against Poisoning Attacks." In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 11, pp. 9524-9532. 2021. LLNL-CONF-814665.
Song, Hoseung, Jayaraman J. Thiagarajan, and Bhavya Kailkhura. "Preventing Failures by Dataset Shift Detection in Safety-Critical Graph Applications," Frontiers in artificial intelligence 4: 60. 2021. LLNL-JRNL-822138.
Thiagarajan, Jayaraman J. “From Predictions to Prescriptive Decisions.” Invited Talk, DNN Next-Gen AI Workshop. LLNL-PRES-825005.
