Skip to content

Instantly share code, notes, and snippets.

View vene's full-sized avatar
🏴
ahoy

Vlad Niculae vene

🏴
ahoy
View GitHub Profile
"""
(C) August 2013, Mathieu Blondel
# License: BSD 3 clause
Custom group support by Vlad Niculae (vlad@vene.ro)
This is a Numba-based reimplementation of the block coordinate descent solver
(without line search) described in the paper:
Block Coordinate Descent Algorithms for Large-scale Sparse Multiclass
@vene
vene / lbfgs_nnls.py
Last active December 10, 2015 11:48 — forked from mblondel/lbfgs_nnls.py
# (C) Mathieu Blondel 2012
import numpy as np
from scipy.optimize import fmin_l_bfgs_b
from sklearn.base import BaseEstimator, RegressorMixin
from sklearn.utils.extmath import safe_sparse_dot
class LbfgsNNLS(BaseEstimator, RegressorMixin):