Miles Cranmer
Miles Cranmer
Rather than requiring a user to set up `extra_sympy_mappings` whenever they define a custom operator, I think the expressions should be evaluated directly in Julia, without needing to convert them....
Have you used PySR in a research paper to discover or rediscover some symbolic model? If so, please add your paper to the following research showcase: https://astroautomata.com/PySR/#/papers You get publicity...
It would be great if PySR would automatically generate plots for discovered expressions, as well as LaTeX of the expression form, during training. Then, one does not need to stop...
This will be primarily tracked in the SymbolicRegression.jl repo: https://github.com/MilesCranmer/SymbolicRegression.jl/issues/89. I think the PySR frontend would need to use a map from variable name to differential operator, so that a...
Julia should print to stdout rather than stderr, as is default for `Pkg.update`, `Pkg.activate`, etc. This can be set with the `io` flag that I pass for these. (stderr prints...
To define a symbolic regression model predict recurrent relations in sequences, I want to create a `PySRSequenceRegressor` (see #88 for scikit-learn API) that will automatically set this up for the...
Apparently binary (number) operators already work: https://github.com/MilesCranmer/SymbolicRegression.jl/issues/58#issuecomment-1003180128, so I would like to implement these as built-in, so that you can simply write "AND" and it will register the corresponding binary...
There are potential problems with only storing Project.toml and not the Manifest.toml file. Simply fixing Project.toml means the Julia dependencies could push patches, and with default PySR settings (`update=True`), there...
Not sure if this is possible in the current setup, but would be nice to have a progress bar display inside the jupyter notebooks. To implement the multi-line progress bar,...
I think something that may give a large algorithmic performance is recursive feature selection. Here is the idea: 1. Hold "next-mutation" feature importances for every equation separately. 2. Start these...