Comments (7)
Hello, Andrew
Thank you for the feedback!
Could you provide an example of incorrect results?
Example of how it supposed to work:
import numpy as np
import pynanoflann
X_index = np.array([[0, 0], [100., 0.], [0., 100.], [100., 100.]])
"""
2 3
|
|
|
|
|
0--------------1
"""
nn = pynanoflann.KDTree(n_neighbors=5, metric='L1', radius=100)
nn.fit(X_index)
query = X_index[:1] # point at (0, 0)
_, indices = nn.radius_neighbors(query, radius=0)
print(indices) # no matches
_, indices = nn.radius_neighbors(query, radius=50)
print(indices) # point at (0, 0)
_, indices = nn.radius_neighbors(query, radius=200-0.01)
print(indices) # points (0, 0), (100, 0), (0, 100); distance to (100, 100) == 200
_, indices = nn.radius_neighbors(query, radius=200+0.01)
print(indices) # all points
Output:
[]
[[0]]
[[0 1 2]]
[[0 1 2 3]]
that is, the results depends on the radius.
from pynanoflann.
Thanks!
I guess my confusion is what the radius parameter in
nn = pynanoflann.KDTree(n_neighbors=5, metric='L1', radius=100)
does.
Edit:
Also curious what the canonical behavior for _, indices = nn.radius_neighbors(query, radius=distance)
is, when we cannot find up to n_neighbors
for a subset of query points within a radius. It seems like in this case the function returns effectively an empty array.
Edit2:
pynanoflann seems to ignore the n_neighbors argument when performing a radius_neighbors query.
from pynanoflann.
For the radius query, there seems to be a minor bug. The python interface is returning square rooted distance, but the radius query takes in a squared query.
from pynanoflann.
For the radius query, there seems to be a minor bug. The python interface is returning square rooted distance, but the radius query takes in a squared query.
@aluo-x that's expected behaviour then, for a L2 query, nanoflann takes in a squared radius and returns a list of squared distances, so the python interface squares the provided euclidean radius as input and then takes the square root of the squared distances to return regular distances.
If you specify a radius from nn = pynanoflann.KDTree(n_neighbors=5, metric='L1', radius=100)
, then this distance 100 will be squared before passing to nanoflann. If, however, you specify a radius from nn.radius_neighbors(query, radius=100)
, then this radius won't be squared before passing to nanoflann, so you'd have to do nn.radius_neighbors(query, radius=100**2)
to get expected behavior.
If that's confusing, I can submit a PR which squares the radius passed to radius_neighbors
too, if @u1234x1234 thinks it's a good idea..
from pynanoflann.
I'm not sure that is the expected behavior. The python interface as of right now uses euclidean distance as function input, but returns squared euclidean distances.
The KDTree construction also seems to ignore the radius argument.
Edit: I mean to say, the function takes in squared euclidean distance, but returns euclidean distance.
from pynanoflann.
Yeah, I agree it is confusing that the constructor radius is the regular distance but the radius_neighbors
radius requires the squared distance, it would probably be good to make those consistent.
from pynanoflann.
Sorry for the confusion, I've made a fix to this bug:
9a93218
You could try new version:
pip install git+https://github.com/u1234x1234/[email protected]
Thank you for the feedback
from pynanoflann.
Related Issues (9)
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from pynanoflann.