8000 [RFC] Minimal scipy version for 1.0 (or 0.26) release · Issue #19705 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]

Skip to content

[RFC] Minimal scipy version for 1.0 (or 0.26) release #19705

@lorentzenchr

Description

@lorentzenchr

Proposal

I'd like to propose to increase the minimal scipy version to 1.0.

SCIPY_MIN_VERSION = '1.0.0'

Reasoning

  1. In case we should release scikit-learn 1.0, it would be a good fit:smirk:
  2. Linear quantile regression [MRG] Add quantile regression #9978 could make it into the next release. It uses scipy.optimize.linprog under the hood. Scipy 1.0.0 has introduced a new solver method="interior-point" which is set as default method. Having it available would help us to avoid to support the "simplex" method in scikit-learn. Note, that scipy v1.3.0 introduced the "revised simplex" method and version 1.5 the "highs**" solvers which are much preferred.
    I think we should avoid the legacy simplex method.
  3. Your reason for scipy 1.0.0.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0