Mapping Machine Learning to Physics (ML2) Proposers Day

Background
The Department of Defense (DoD) is initiating the Machine Learning to Physics (ML2P) program to address the challenges of deploying machine learning (ML) in resource-constrained battlefield environments. As ML technology evolves, its energy demands increase, necessitating a sustainable approach to ML that balances power consumption with performance. The ML2P program aims to enhance the military's capability to utilize ML effectively on the battlefield by prioritizing energy efficiency and optimizing power usage throughout the ML lifecycle.

Work Details
The ML2P program will focus on several key technical goals:

1. Develop multi-objective optimization functions that balance power consumption with performance metrics across various hardware architectures.

2. Map ML efficiency directly to physics using precise Joule measurements for accurate power and performance predictions.

3. Construct energy-aware ML models that optimize power (Joules) and performance metrics (Accuracy, Precision, Recall, F1 score) for specific tasks and hardware configurations.

4. Explore interactions between local optimizations through Energy Semantics of ML (ES-ML) to solve complex energy-aware ML optimization problems.

Period of Performance
The contract's period of performance is not explicitly stated but involves ongoing research and development efforts aligned with the objectives of the ML2P program.

Place of Performance
The primary location for the Proposers Day event is at the Executive Conference Center, located at 4075 Wilson Blvd, Arlington, Virginia 22203.