Using neural networks to quickly approximate the results of computationally-intensive simulations A collaborative research effort finds that design teams can now use neural networks to evaluate fluid dynamics issues and dramatically increase the number of design options they consider. ABSTRACT This paper explores a collaborative research effort that demonstrated the power of artificial intelligence to stand in for the traditional simulation and modeling workloads that typically run in high performance computing systems. In this project, our research team determined that the use of trained neural networks can allow design teams to quickly evaluate the viability of many different design alternatives in comparison to running computationally-intensive simulations for each alternativ

Join for free to read