I am still looking into this - it seems to be related to the python-numpy-mkl library in arch. I installed the regular python-numpy package and it has removed all the issues.

So i built python-graph-tool-git-2.29.r34.gcdcddcc4-1 from the AUR and I can confirm that I am still getting the NaNs when printing x using the example above. If you repeat the full computation, do you never see any NaNs?

Hi Tiago,

Many thanks, I will compile the current git version from the AUR this afternoon and let you know.

I should also note, that approximately once out of every five or so times, I will get the result with NaNs.

Hi Tiago,

Many thanks for your reply! Please see the bellow complete example -

```
import graph_tool.all as gt
import numpy as np
g = gt.collection.data["polblogs"]
g = gt.GraphView(g, vfilt=gt.label_largest_component(g))
w = g.new_edge_property("double")
w.a = np.random.random(len(w.a)) * 42
ee, x = gt.eigenvector(g, w)
print(x.a)
```

Hi There,

On the latest build of graph-tool 2.29 from the AUR on Arch, I am getting Nans whenever I try to run the Eigenvector function on any I graph I try. Not sure if this is expected behaviour or not.

As a demonstration, the example code from the documentation will return Nans -

```
g = gt.GraphView(g, vfilt=gt.label_largest_component(g))
w = g.new_edge_property("double")
w.a = np.random.random(len(w.a)) * 42
ee, x = gt.eigenvector(g, w) ```
Will result in the following:
PropertyArray([nan, nan, 0., ..., 0., nan, 0.])
```