Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9199

tttthomasssss · 2017-06-22T12:53:11Z

Description

Fitting a NearestNeighbors model fails when a) the distance metric used is a callable and b) the input to the NearestNeighbors model is sparse.

Steps/Code to Reproduce

from scipy import sparse
from sklearn.neighbors import NearestNeighbors

def sparse_metric(x, y): # Some metric accepting sparse input
    return x.count_nonzero() / y.count_nonzero()

A = sparse.random(10, 5, density=0.3, format='csr')

nn = NearestNeighbors(algorithm='brute', metric=sparse_metric).fit(A)

Expected Results

No error is thrown when passing a callable as metric with sparse input

Actual Results

ValueError                                Traceback (most recent call last)
<ipython-input-2-a9d2fd7f843b> in <module>()
      7 A = sparse.random(10, 5, density=0.3, format='csr')
      8 
----> 9 nn = NearestNeighbors(algorithm='brute', metric=sparse_metric).fit(A)

/Volumes/LocalDataHD/thk22/.virtualenvs/nlpy3/lib/python3.5/site-packages/sklearn/neighbors/base.py in fit(self, X, y)
    797             or [n_samples, n_samples] if metric='precomputed'.
    798         """
--> 799         return self._fit(X)

/Volumes/LocalDataHD/thk22/.virtualenvs/nlpy3/lib/python3.5/site-packages/sklearn/neighbors/base.py in _fit(self, X)
    213             if self.effective_metric_ not in VALID_METRICS_SPARSE['brute']:
    214                 raise ValueError("metric '%s' not valid for sparse input"
--> 215                                  % self.effective_metric_)
    216             self._fit_X = X.copy()
    217             self._tree = None

ValueError: metric '<function sparse_metric at 0x1097d0378>' not valid for sparse input

Some Analysis/Wild Speculation

The problem seems to come from the fact that in the case of sparse input, it is only checked whether the given metric is in the list of metrics accepting sparse input, but no check is made whether the given metric is a string or a callable: https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/neighbors/base.py#L210

Versions

Darwin-15.6.0-x86_64-i386-64bit
Python 3.5.1 (default, Dec  8 2015, 06:00:08) 
[GCC 4.2.1 Compatible Apple LLVM 6.0 (clang-600.0.57)]
NumPy 1.12.1
SciPy 0.19.0
Scikit-Learn 0.18.2

The text was updated successfully, but these errors were encountered:

jnothman · 2017-06-22T13:05:56Z

I think this may be fixed in master due to #9145

jnothman · 2017-06-22T13:06:00Z

B

jnothman · 2017-06-22T13:06:09Z

Please check

tttthomasssss · 2017-06-22T15:26:25Z

I just cloned master and it results in the same error. The codeline I referenced above was actually from master as well.

ValueError                                Traceback (most recent call last)
<ipython-input-2-a9d2fd7f843b> in <module>()
      7 A = sparse.random(10, 5, density=0.3, format='csr')
      8 
----> 9 nn = NearestNeighbors(algorithm='brute', metric=sparse_metric).fit(A)

~/DevSandbox/InfiniteSandbox/scikit-learn/sklearn/neighbors/base.py in fit(self, X, y)
    801             or [n_samples, n_samples] if metric='precomputed'.
    802         """
--> 803         return self._fit(X)

~/DevSandbox/InfiniteSandbox/scikit-learn/sklearn/neighbors/base.py in _fit(self, X)
    214             if self.effective_metric_ not in VALID_METRICS_SPARSE['brute']:
    215                 raise ValueError("metric '%s' not valid for sparse input"
--> 216                                  % self.effective_metric_)
    217             self._fit_X = X.copy()
    218             self._tree = None

ValueError: metric '<function sparse_metric at 0x1122a57b8>' not valid for sparse input

Versions

Darwin-15.6.0-x86_64-i386-64bit
Python 3.5.1 (default, Dec 8 2015, 06:00:08)
[GCC 4.2.1 Compatible Apple LLVM 6.0 (clang-600.0.57)]
NumPy 1.13.0
SciPy 0.19.0
Scikit-Learn 0.19.dev0

tttthomasssss · 2017-06-22T20:11:52Z

A very simple fix would be to change Line 214 in sklearn/neighbors/base.py from

if self.effective_metric_ not in VALID_METRICS_SPARSE['brute']:

to

if not callable(self.effective_metric_) and self.effective_metric_ not in VALID_METRICS_SPARSE['brute']:

jnothman · 2017-06-22T22:25:12Z

:( Sorry we didn't fix this in the previous patch. Pull request welcome, with a test.

…

On 23 June 2017 at 06:12, Thomas ***@***.***> wrote: A very simple fix would be to change Line 214 <https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/neighbors/base.py#L214> in sklearn/neighbors/base.py <https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/neighbors/base.py> from if self.effective_metric_ not in VALID_METRICS_SPARSE['brute']: to if not callable(self.effective_metric_) and self.effective_metric_ not in VALID_METRICS_SPARSE['brute']: — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#9199 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz61byWrAlsjvZAS7YUkKtVfq08pZ9ks5sGsqRgaJpZM4OCQLN> .

tttthomasssss · 2017-06-22T22:55:53Z

Cool, I'll submit a PR along with a test for it.

taineleau-zz · 2017-08-15T21:57:32Z

@tttthomasssss have you fixed this bug?

jnothman · 2017-08-15T23:08:16Z

No pull request yet

tttthomasssss · 2017-08-16T08:40:01Z

@taineleau @jnothman thanks for the reminder and apologies for the delay!

I have a fix for the problem per se, however haven't gotten round to write an appropriate test for it.

I will submit a PR within the next few days.

taineleau-zz · 2017-08-16T23:16:47Z

Hope you fix it soon! Thanks!!!!!

…

On Wed, Aug 16, 2017 at 4:40 AM, Thomas ***@***.***> wrote: @taineleau <https://github.com/taineleau> @jnothman <https://github.com/jnothman> thanks for the reminder and apologies for the delay! I have a fix for the problem *per se*, however haven't gotten round to write an appropriate test for it. I will submit a PR within the next few days. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#9199 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGKWDmMRQZIUF4kMa2iVjyBxT8997sxFks5sYqsEgaJpZM4OCQLN> .

-- Danlu Chen Homepage: https://taineleau.me

…ghbors model fails with sparse input and a callable as metric

…rse input and a callable as metric (#9579)

…ils with sparse input and a callable as metric (scikit-learn#9579)

jnothman closed this as completed Jun 22, 2017

jnothman reopened this Jun 22, 2017

tttthomasssss added a commit to tttthomasssss/scikit-learn that referenced this issue Aug 17, 2017

proposed fix (incl test) for scikit-learn#9199 - Fitting a NearestNei…

2e315eb

…ghbors model fails with sparse input and a callable as metric

tttthomasssss mentioned this issue Aug 17, 2017

[MRG+1] Fixes #9199 - Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9579

Merged

TomDLT closed this as completed in #9579 Dec 6, 2017

TomDLT pushed a commit that referenced this issue Dec 6, 2017

[MRG+1] Fixes #9199 - Fitting a NearestNeighbors model fails with spa…

ece341c

…rse input and a callable as metric (#9579)

tttthomasssss added a commit to tttthomasssss/scikit-learn that referenced this issue Dec 14, 2017

Added whats new entry for issue scikit-learn#9199 (pr scikit-learn#9579)

651498e

jwjohnson314 pushed a commit to jwjohnson314/scikit-learn that referenced this issue Dec 18, 2017

[MRG+1] Fixes scikit-learn#9199 - Fitting a NearestNeighbors model fa…

8a66640

…ils with sparse input and a callable as metric (scikit-learn#9579)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9199

Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9199

tttthomasssss commented Jun 22, 2017 •

edited

Loading

jnothman commented Jun 22, 2017

jnothman commented Jun 22, 2017

jnothman commented Jun 22, 2017

tttthomasssss commented Jun 22, 2017

tttthomasssss commented Jun 22, 2017

jnothman commented Jun 22, 2017 via email

tttthomasssss commented Jun 22, 2017

taineleau-zz commented Aug 15, 2017

jnothman commented Aug 15, 2017

tttthomasssss commented Aug 16, 2017

taineleau-zz commented Aug 16, 2017 via email

Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9199

Fitting a NearestNeighbors model fails with sparse input and a callable as metric #9199

Comments

tttthomasssss commented Jun 22, 2017 • edited Loading

Description

Steps/Code to Reproduce

Expected Results

Actual Results

Some Analysis/Wild Speculation

Versions

jnothman commented Jun 22, 2017

jnothman commented Jun 22, 2017

jnothman commented Jun 22, 2017

tttthomasssss commented Jun 22, 2017

Versions

tttthomasssss commented Jun 22, 2017

jnothman commented Jun 22, 2017 via email

tttthomasssss commented Jun 22, 2017

taineleau-zz commented Aug 15, 2017

jnothman commented Aug 15, 2017

tttthomasssss commented Aug 16, 2017

taineleau-zz commented Aug 16, 2017 via email

tttthomasssss commented Jun 22, 2017 •

edited

Loading