- This is a symbolic regression algorithm with Gene Expression Programming as a role model. (the operators changed - for experimental reasons)
- At a certain amount of expression, the list iterations within the genetic operators produce a lot of computational overhead
- Here, we project these into a vector space and use simple vector operations
- Velocity check: 1000 for 1000 epochs: approx 4:30m python, 0:48m julia
- Inner approximation of the constants with perturbation and the application of the gene fusion operator
- Vectorization cost evaluation
- Vectorization genetic operators
- Random application of genetic operators
- Conversion to Julia - (Needs to be tested/debugged)
- Implementing of learnable weights for the genetic operators
- Implementation of an attention-based mutation operator
- Automatic gene-len scaling
#Remarks
- Please add some comments