glum
glum copied to clipboard
Quantco
High performance Python GLMs with all the features!
The [r2_score](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.r2_score.html) method in sklearn has a force_finite flag which defaults to True in order to avoid infinite and NaN values when the TSS happens to be 0. The analogous...
I'm trying to fit Tweedie and Poisson based GLMs to data that is heavy in 0 values. When fitting a GLM with a log link, the glum code tries to...
We should update the benchmarks to reflect the latest developments in other libraries.
Instead of storing the entire Hessian matrix, `diag_fisher=True` calculates Hessian rows _only when they must be used_ (i.e. in the relevant step of coordinate descent). It accomplishes this by storing...
When `use_sparse_hessian=True`, `state.hessian` is stored as a sparse COO matrix. Partially addresses Issue #485 -- the dataset with 4M+ columns can now run without a memory error.
Recently, our glmnet benchmark testing has stopped working in the CI. The error looks similar to [this issue](https://github.com/rpy2/rpy2/issues/874). In #537, I disabled the glmnet benchmark test so that the CI...
The labels are actually there if you squint because they're in a slightly different shade of black than the background. Maybe there's an easy/common way of handling this?
Would there be any interest in the addition of the Cox Proportional Hazards loss for modeling survival data?
I would be very nice to also fit completely unpenalized GLMs (as in base R glm). For categorical features, one then needs to drop one level, called reference or base...
Over at quantcore.learn, we use [asv](https://github.com/airspeed-velocity/asv) to perform automatic performance benchmarks as part of our CI pipeline. We have a dedicated machine at [kimsufi](https://www.kimsufi.com/en/) so that we're running these tests...
