首页>
外国专利>
NEURAL GRAPH TRANSFORMER NETWORK FORCE FIELD FOR THE PREDICTION OF ATOMIC FORCES AND ENERGIES IN MOLECULAR DYNAMICS SIMULATIONS
NEURAL GRAPH TRANSFORMER NETWORK FORCE FIELD FOR THE PREDICTION OF ATOMIC FORCES AND ENERGIES IN MOLECULAR DYNAMICS SIMULATIONS
展开▼
机译:用于预测分子动力学模拟原子力和能量的神经图变压器网络力领域
展开▼
页面导航
摘要
著录项
相似文献
摘要
A simulation involves converting a molecular dynamics snapshot of the elements within the multi-element system into a graph with atoms as nodes of the graph, defining a matrix so that each column of the matrix represents a node in the graph, defining a distance matrix according to a set of relative positions of the respective atoms, iterating over the GTFF using an attention mechanism which acts on the matrix and is extended by the inclusion of the distance matrix in order to transfer the hidden state from a current layer of the GTFF to a next layer of the GTFF, executing a Combining across the columns of the matrix to produce a scalar molecular energy, running back through the GTFF, iteratively calculating derivatives at each of the layers of the GTFF to compute a prediction of the force acting on each atom, and returning the prediction of the force acting on every atom t.
展开▼