2019 NOVEL APPROACHES TO LARGE SCALE SCIENTIFIC COMPUTING
Between TensorFlow, particularly the TPU chips, limited precision arithmetic, more mathematics for program construction and data modeling, I believe there are opportunities for dramatic improvements for some grand challenge compute problems.
2018- MACHINE LEARNING
Applications of Machine Learning to simulation Physics Equations
Use of TensorFlow for High Performance Computing
Interpretability of ML models
2018- LIMITED PRECISION AND NUMBER FORMATS FOR SKA
The role of floating point number formats and precision for SKA compute pipelines. We expect this can dramatically reduce the requirement for bandwidth.
2017- SOFTWARE FOR NEW MEMORY TECHNOLOGIES
Concepts in programming languages and runtimes related to persistent and hybrid memory.
Automatic and intelligent control of caches.
2014- HARDWARE AND SOFTWARE FOR PARALLEL PROGRAMMING
Geometric underpinnings of efficient parallel algorithms, such as network topologies and data organization