Olivier Grisel

Results 117 issues of Olivier Grisel

Should a dedicated API/column metadata to efficiently support sparse columns be part of the spec? ## Context It can be the case than a given column has more more than...

enhancement

Currently we fix `n_iter`. I think we should instead implement some kind of stochastic convergence criterion (either on the training data or on held out validation split) and instead configure...

Unfortunately, I cannot see the `soda-inria/hazardous` repo from my codecov account at the moment (even after granting the permissions to codecov to access the `soda-inria` organization).

vtune can be installed from the oneAPI basekit. Here is the command to record GPU statistics from the benchmark: ``` vtune -collect gpu-hotspots -- python kmeans.py --n-data-samples 5000000 --engine-providers sklearn-numba-dpex...

https://github.com/IntelPython/dpcpp-llvm-spirv should be available from the `dppy/label/dev` conda channel. It should be much lighter weight than the `"intel::dpcpp_linux-64"` we currently use as a dependency for `sklearn_numba_dpex`.

At the moment the only way to select a given device (cpu, gpu, with level zero or OpenCL or default host device) is via the `SYCL_DEVICE_FILTER` environment variable. It would...

Note: this PR uses `xp.linalg.solve` without being able to pass `assume_a="pos"` as is [possible in scipy](https://docs.scipy.org/doc/scipy/reference/generated/scipy.linalg.solve.html). As a result, the precise nature of the square matrix decomposition used to solver...

module:linear_model
module:utils
Array API

The goal of this helper would be to make it possible to have education examples that download and cache datafiles that are then manually loaded with functions such as `pandas.read_csv`,...

module:datasets

This PR fixes two usability problems with the new `TunedThresholdClassifierCV` when using it with invalid values for the `scoring` parameter: - the first case, is passing a scoring name or...

Bug
module:model_selection
To backport

```python >>> import dask.array as da >>> da.float32 Traceback (most recent call last): Cell In[5], line 1 da.float32 AttributeError: module 'dask.array' has no attribute 'float32' >>> da.isdtype Traceback (most recent...

needs attention
needs triage