Added
- New module "map" to construct a matrix Taylor map for the given system of ODEs
- A module to numerically solve systems of ODEs – "tm_solver" based on the "map" module.
- Examples to illustrate the use of "tm_solver" in the numerical solution of example ODEs (forward problem):
- Examples of ODE systems (Lotka-Volterra system, Robertson system, generalized Lotka-Volterra system).
- Example of ODE system with symbolic parameters on the right-hand side
- A "learn" module example based on the Lotka-Volterra system to show the ability of PNN
to learn system dynamics from data (inverse problem)
and to demonstrate better training results of pre-initialized PNN
over PNN without initialization
Changed
- Simplified tmflow imports:
from tmflow.learn import KroneckerPolynomialLayer instead of
from tmflow.learn.kronecker_polynomial_layer import KroneckerPolynomialLayer