jakevdp

import numpy as np
from scipy.sparse.linalg import eigs

N = 6
k = 2

# with this random seed, I get a memory error on the third iteration below
np.random.seed(2301)
A = np.random.random((N,N))

jakevdp

code demonstrating the problem seen in issue #365

to run the example:
  tar -zxvf data.tgz
  python test.py

jakevdp

from time import time
import numpy as np
from scipy.sparse import spdiags, issparse, dia_matrix
from scipy.sparse.linalg import factorized
from scipy import linalg as splinalg

class BandedMatrix(object):
    def __init__(self, data, lu=None):
        if issparse(data):
            if lu:

jakevdp

This includes a test of the new GMM routines in https://github.com/bthirion/scikit-learn/tree/gmm-fixes

By changing the line

GMM = mixture.GMM

at the top of the file, we can plot the BIC and AIC for each variant of GMM. Standard GMM works beautifully: it settles in on 3 components, which are a good description of the data. DPGMM and VBGMM produce some unexpected results.

jakevdp

This is the outline of a framework that will allow general distance metrics to be incorporated into scikit-learn BallTree. The idea is that we need a fast way to compute the distance between two points under a given metric. In the basic framework here, this involves creating an object which exposes C-pointers to a function and a parameter structure so that the distance function can be called from either python or directly from cython with no python overhead.

jakevdp

SPHINXBUILD   = sphinx-build
BUILDDIR      = _build
SPHINXOPTS   = -d $(BUILDDIR)/doctrees .

all: html

html:
	$(SPHINXBUILD) -b html $(SPHINXOPTS) $(BUILDDIR)/html
	@echo
	@echo "Build finished. The HTML pages are in $(BUILDDIR)/html."

jakevdp

import warnings
from sklearn import datasets
from sklearn.neighbors import NearestNeighbors
import numpy as np

n_points = 1000
n_neighbors = 10
out_dim = 2

n_trials = 100

jakevdp

"""
This file has an example function, with a documentation string which should
serve as a template for scikit-learn docstrings.
"""

def sklearn_template(X, y, a=1, flag=True, f=None, **kwargs):
    """This is where a short one-line description goes

    This is where a longer, multi-line description goes.  It's not
    required, but might be helpful if more information is needed.

jakevdp

import numpy as np
from matplotlib import pyplot as plt
from matplotlib import animation

# First set up the figure, the axis, and the plot element we want to animate
fig = plt.figure()
ax = fig.add_subplot(111, xlim=(0, 2), ylim=(-2, 2))
line, = ax.plot([], [], lw=2)

# initialization function: plot the background of each frame

jakevdp

Numba Ball Tree

This is a quick attempt at writing a ball tree for nearest neighbor searches using numba. I've included a pure python version, and a version with numba jit decorators. Because class support in numba is not yet complete, all the code is factored out to stand-alone functions in the numba version. The resulting code produced by numba is about ~10 times slower than the cython ball tree in scikit-learn. My guess is that part of this stems from lack of inlining in numba, while the rest is due to some sort of overhead