Estimating the Jacobian matrix of an unknown multivariate function from sample values by means of a neural network

Frédéric Latrémolière, Sadananda Narayanappa, Petr Vojtěchovský

We describe, implement and test a novel method for training neural networks to estimate the Jacobian matrix $J$ of an unknown multivariate function $F$. The training set is constru…