Introduction
This bi-monthly seminar series explores real-world applications of physics-informed machine learning (Φ-ML) methods to the engineering practice. They cover a wide range of topics, offering a cross-sectional view of the state of the art on Φ-ML research, worldwide.
Participants have the opportunity to hear from leading researchers and learn about the latest developments in this emerging field. These seminars also offer the chance to identify and spark collaboration opportunities.
About the event
Despite the relevance of uncertainty in computing systems, existing computer architectures neither have support for representing uncertainty, nor for arithmetic on uncertain values. We present Laplace, a processor microarchitecture for tracking machine representations of probability distributions paired with integer and floating-point architectural state.
We present two new methods for in-processor distribution representations which are approximations of probability distributions just as floating-point number representations are approximations of real-valued numbers. Laplace executes unmodified RISC-V binaries and can track uncertainty through them.
We evaluate the accuracy and performance of Laplace using a suite of 21 benchmarks from a range of domains from variational quantum algorithms and sensor data processing to materials properties modeling. Laplace is able to achieve the same accuracy as Monte Carlo on the benchmarks, using 1350 times fewer instructions on average and up to 15750 times fewer instructions when using larger in-processor representations.
Compared to state-of-the-art alternatives to Monte Carlo, Laplace achieves a 62% improvement versus PaCAL and a 6.3-fold improvement compared to the method used by the NIST Uncertainty Machine, quantified by the Wasserstein distance to Monte Carlo. Unlike existing methods which require software to be rewritten in a domain-specific language or which may require extensive source-level changes, Laplace achieves all of these benefits while requiring no changes to binaries in order to track uncertainty through them, with only minimal changes required to get uncertainty information into the microarchitecture. An implementation of Laplace is now part of a commercial product.
Watch on demand:
