installing from pypi or pip about annoy HOT 8 CLOSED

1 error generated.
error: command 'gcc' failed with exit status 1

from annoy.

Did you install Boost?

I'm planning to drop the dependency on Boost fairly soon. It's been causing a lot of issues for people.

http://www.pyimagesearch.com/2015/04/27/installing-boost-and-boost-python-on-osx-with-homebrew/

from annoy.

I remember liking boost a lot when I was a quant on a trading desk! I'll
install it.

I'm trying to use Annoy instead of a BallTree or KDtree which takes too
long to construct and query: 2.5 million documents TFIDF vectorized.

The goal would be to query 100,000 against 2.5 in a reasonable amount of
time.

Would you have any insight?

• Ajay
  ᐧ

On Tue, Apr 28, 2015 at 3:59 PM, Erik Bernhardsson <[email protected]

wrote:

Did you install Boost?

I'm planning to drop the dependency on Boost fairly soon. It's been
causing a lot of issues for people.

http://www.pyimagesearch.com/2015/04/27/installing-boost-and-boost-python-on-osx-with-homebrew/

—
Reply to this email directly or view it on GitHub
#49 (comment).

Github https://github.com/asharma567/ | LinkedIn
http://www.linkedin.com/in/mrasharma/

from annoy.

If you come back in a few days the Boost dependency should be dropped.

Yes you can do that. Note however for TF-IDF if the number of terms is high the nearest neighbor quality might not be perfect. You should probably do a dimensionality reduction (using matrix factorization etc) to a representation with 10-100 dimensions at most.

from annoy.

Interesting, I'll try that approach -- Thanks!
ᐧ

On Wed, Apr 29, 2015 at 11:14 AM, Erik Bernhardsson <
[email protected]> wrote:

If you come back in a few days the Boost dependency should be dropped.

Yes you can do that. Note however for TF-IDF if the number of terms is
high the nearest neighbor quality might not be perfect. You should probably
do a dimensionality reduction (using matrix factorization etc) to a
representation with 10-100 dimensions at most.

—
Reply to this email directly or view it on GitHub
#49 (comment).

Github https://github.com/asharma567/ | LinkedIn
http://www.linkedin.com/in/mrasharma/

from annoy.

Can you try to install from PyPI now? The newest version (1.1.0) should not depend on Boost (thanks to @dvenum)

from annoy.

It's work for me. Debian 8.

from annoy.

works pypi works : OSX

On Wed, Apr 29, 2015 at 6:03 PM, Erik Bernhardsson <[email protected]

wrote:

Closed #49 #49.

—
Reply to this email directly or view it on GitHub
#49 (comment).

Github https://github.com/asharma567/ | LinkedIn
http://www.linkedin.com/in/mrasharma/

from annoy.