NASA Background: NASA uses the vantage point of space to increase our understanding of our home planet, improve lives, and safeguard our future. The end-to-end capability of NASA Earth Science Division – from developing new technology to delivering actionable science –makes it unique among U.S. government science agencies. NASA’s recent Earth Science to Action strategy notes that “the reliance of governments, society and the economy on environmental information has increased significantly, which translates into demand for more sophisticated, more accurate, more trustworthy, and more actionable information at the fingertips of decision-makers, policymakers, and action takers. This also requires more complex science questions to be addressed, to better understand the Earth-human system.”
Challenge Background. NASA, however, faces challenges in speed or accuracy with classical models and compute systems. Recent advances in innovative and unconventional computing capability are beginning to offer new opportunities. Quantum computers offer a fundamentally different paradigm of computation and can solve certain classes of problems exponentially faster than their classical counterparts. However, not all types of calculations will achieve this improvement in speed. Likewise, quantum machine learning offers the potential to reduce required training data or produce more accurate models by adding quantum layers to neural nets.
The emerging field of neuromorphic, or brain-inspired, computing implements spiking recurrent neural network computation and holds significant promise for algorithm development optimized for high-speed, low power. In-memory computing bypasses the von Neumann bottleneck, saving time and energy for data-heavy processes like AI training and inferencing by putting compute and memory side by side.
Previous Challenge Baselines. A 2024 IEEE challenge asked participants to combine multi-source data to classify flood water surface extent. Participants were evaluated based on a binary water/no-water classification and resulting F1 score. Results of this previous challenge may be built upon using Quantum Machine Learning or other unconventional computing methodologies. Data from this challenge may be used as test input data for the challenge, but participants may also bring their own data based on current research or applicability to solution design and feasibility. More information can be found in the Resources section.
