Miles Cranmer

Results 287 issues of Miles Cranmer

The latest Optim.jl generates warnings when NaN is encountered during optimization: https://github.com/JuliaNLSolvers/Optim.jl/releases/tag/v1.7.7 This results in PySR & SymbolicRegression.jl generating tons of such warnings... So would be great if this could...

### Feature Request Rather than making the user define custom `extra_sympy_mappings`, we could potentially take advantage of [SymPy.jl](https://github.com/JuliaPy/SymPy.jl) and automatically generate the sympy operators. Furthermore, and this would be more...

enhancement
priority: low

This should hopefully provide the debugging info up front so users don't run into #362. For every user who posts a bug report there are probably 10 who simply give...

Rather than `@pysr-0.14.2` as the Julia environment, this would make it `@pysr-0.14.2-py-3.11-jl-1.9`. Thus the user would need to re-install for each combination of Python and Julia they wish to use...

### Feature Request It might be useful to explore adding expression transformation rules which do not simplify the expression, but rather transform it to a more complex expression which is...

enhancement

### Feature Request Should use Measurements.jl on the Julia side, and allow [`uncertainties`](https://pythonhosted.org/uncertainties/) array as input. Loss function maybe could be Gaussian-Gaussian likelihood? Or KL divergence?

enhancement

This is long-overdue but I finally finished a methods paper describing the algorithm in PySR and SymbolicRegression.jl. You can find it here: https://github.com/MilesCranmer/pysr_paper and the arXiv here: https://arxiv.org/abs/2305.01582. I consider...

documentation

Right now it seems impossible to stop early within Jupyter notebooks. If one enters a search that takes a long time, the only way to quit is to restart the...

bug
priority: high

The current list of options is way too long to be understood by a user. I think a refactoring should be done where new objects are used to hierarchically define...

documentation
priority: mid

PySRRegressor currently has 74 parameters. That is way too many for a single method. It can be hard to figure out what a single parameter from the long list actually...