Professor Csaba Andras Moritz and his Ph.D. student Sourabh Kulkarni received some healthy publicity on their probabilistic artificial intelligence (AI) work for COVID-19 modeling, as funded by Facebook’s AI group. A paper from the UMass researchers, collaborating with those from Facebook and Graphcore, demonstrated how essential COVID-19 modeling and analysis, using an Approximate Bayesian Computation (ABC) algorithm, can be massively accelerated with emerging hardware.
Moritz and his team used the COVID-19 analysis to demonstrate their groundbreaking research in their previously published paper (“Accelerating Simulation-based Inference with Emerging AI Hardware”), authored by Kulkarni, Moritz, A. Tsyplikhin, and Mario Michael Krell for the Institute of Electrical and Electronics Engineers International (IEEE) Conference on Rebooting Computing in 2020.
As the IEEE paper explained, “In this work, we explore hardware accelerated simulation-based inference over probabilistic models by combining a massively parallelized ABC inference algorithm with the cutting-edge AI chip solutions that are uniquely suited for this purpose.”
The results of the collaborative research indicated an impressive 30 times speedup on IPUs compared with CPUs and a significant 7.5 times speedup compared with GPUs.
Subsequently, the researchers used their new COVID-19 analysis as a proof-of-concept model. As the Graphcore paper, written by Krell, explained, “As many of us now know, to tackle a pandemic effectively, it is important to understand how infections spread in a population and how different interventions can impact the spread.
Krell explained that parameters which are particularly useful for this purpose include the infection rate, recovery rate, positive test rate, fatality rate, and testing protocol effectiveness.
“Since the spread of a virus is not a deterministic process at a macroscopic level, researchers are interested in the distribution of these parameters rather than point estimates,” Krell wrote. “Knowing these distributions enables us to find out where significant differences can be observed in contrast to spurious deviations.”
According to the IEEE conference paper written by Moritz, his student Kulkarni, and his fellow researchers, developing models of natural phenomena by capturing their underlying complex interactions is a core tenet of various scientific disciplines. These models can help in understanding the natural processes being studied.
Moritz and colleagues added that one key challenge in this pursuit has been to enable statistical inference over these models, which would allow these simulation-based models to learn from real-world observations.
“Recent efforts, such as ABC, show promise in performing a new kind of inference to leverage these models,” the research team concluded. “While the scope of applicability of these inference algorithms is limited by the capabilities of contemporary computational hardware, they show potential of being greatly accelerated.”
According to the research team, “As a proof-of-concept, we demonstrate inference over a probabilistic epidemiology model used to predict the spread of COVID-19. Two hardware acceleration platforms are compared - the Tesla V100 GPU and the Graphcore Mark1 IPU. Our results show that, while both of these platforms significantly outperform multi-core CPUs, the Graphcore Mark1 IPUs are 7.5 times faster than the Tesla V100 GPUs for this workload.” (February 2021)