首页|Princeton University Reports Findings in Machine Learning [Qu asi- Classical Trajectory Calculation of Rate Constants Using an Ab Initio Traine d Machine Learning Model (aML-MD) with Multifidelity Data]
Princeton University Reports Findings in Machine Learning [Qu asi- Classical Trajectory Calculation of Rate Constants Using an Ab Initio Traine d Machine Learning Model (aML-MD) with Multifidelity Data]
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – New research on Machine Learning is th e subject of a report. According to news reporting out of Princeton, New Jersey, by NewsRx editors, research stated, “Machine learning (ML) provides a great opp ortunity for the construction of models with improved accuracy in classical mole cular dynamics (MD). However, the accuracy of a ML trained model is limited by t he quality and quantity of the training data.” Our news journalists obtained a quote from the research from Princeton Universit y, “Generating large sets of accurate ab initio training data can require signif icant computational resources. Furthermore, inconsistent or incompatible data wi th different accuracies obtained using different methods may lead to biased or u nreliable ML models that do not accurately represent the underlying physics. Rec ently, transfer learning showed its potential for avoiding these problems as wel l as for improving the accuracy efficiency, and generalization of ML models usi ng multifidelity data. In this work, ab initio trained MLbased MD (aML-MD) mode ls are developed through transfer learning using DFT and multireference data fro m multiple sources with varying accuracy within the Deep Potential MD framework. The accuracy of the force field is demonstrated by calculating rate constants f or the H + HO -> H + O reaction using quasi-classical tr ajectories. We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than fi ve times compared to the use of more expensive quantum chemistry training data s ets.”
PrincetonNew JerseyUnited StatesNo rth and Central AmericaCyborgsEmerging TechnologiesMachine Learning
NETL
NSTL
