How to optimize constants in a model？
wangsy15 opened this issue · comments
How to incorporate a local optimizater into GP symbolic formula dedicated to constant number optimization, or use other techniques. For example, i have a formula
d(v1)/dt=add(v1, add(add(v2, 5.426), mul(v1, 3.266))),
it is difficult to learn its coefficient because it is not an integer.
How to check all the constants contained in each formula tree, namely count their number and record their position in the tree？
I used GP in DEAP to solve the symbolic regression problem, specifically trying to fit the data with the most accurate symbolic formula. I would like to ask how to find all the constants contained in each tree if the best formula obtained is represented as a binary tree, i.e. calculate their number n and record their position in the tree. For example, if the best formula is
d(v1)/dt=add(v1, add(add(v2, 5.426), mul(v1, 3.266))),
is it possible to find 5.426 and 3.266 and give their positions in the tree. Is there a function in deap that can implement this requirement?
Or I would like to ask how to optimise the parameters in the equations during genetic programming. Something like expressing the equation in the form Y=FP, with F being the matrix containing the operators and P being the parameter matrix. How can P be optimised while keeping the structure of the equation constant (i.e. F is constant).