Tag
This paper establishes an exact correspondence between neural network training and Hamilton-Jacobi initial-value problems, unifying deep learning architectures through a deformation parameter.