I tried the programs in the tutorial and also the code of the guy that reported the other issue and i get the following error:

/usr/local/lib/python3.7/site-packages/hypernetx/drawing/rubber_band.py:97: FutureWarning: arrays to stack must be passed as a "sequence" type such as list or tuple. Support for non-sequence iterables such as generators is deprecated as of NumPy 1.16 and will raise an error in the future.

I don't know if the problem is that I use python3

Here is a new PR for the weighted hypernetworks code, with all the new commits in a separate branch. Here is the original comment for reference:

I thought I would try to make it a bit easier to create a weighted hypergraph, using a similar syntax to that in networkx. Now you can add weighted edges like this:

w = 0.019281
G = hnx.Hypergraph()
G.add_edge(("A", "B"), weight=w)

I've also added support for calculating the weighted ~~node~~ adjacency matrices:

w0 = 0.98279384
w1 = 0.2309
G = hnx.Hypergraph()
e0 = [("A", "B"), ("B", "C")]
e1 = [("A", "B", "C")]
G.add_edges_from(e0, weight=w0)
G.add_edges_from(e1, weight=w1)
	
print(G.adjacency_matrix())
print(G.adjacency_matrix(weight="weight"))
print(G.edge_adjacency_matrix(weight="weight"))

~~I have an issue with the edge adjacency matrix, because I would ordinarily calculate it with the following matrix formula: \sqrt(W) * M^T * M * \sqrt(W), but this won't currently work as edge_adjacency_matrix calls __incidence_to_adjacency, the same method as is called by adjacency_matrix. I would be grateful for your thoughts on this, I've currently left the weighted edge adjacency to throw a NotImplementedError.~~

I have modified the code to do what I commented in the original PR. When the weighted adjacency matrix is requested, M.dot(weight_matrix) is passed into __incidence_to_adjacency so both adjacency matrix functions work.

All of these changes play quite nicely with the existing code, but I wanted to submit them for your consideration before doing too much more.

Many thanks. :)

Hi,

I thought I would try to make it a bit easier to create a weighted hypergraph, using a similar syntax to that in networkx. Now you can add weighted edges like this:

w = 0.019281
G = hnx.Hypergraph()
G.add_edge(("A", "B"), weight=w)

I've also added support for calculating the weighted node adjacency matrix:

w0 = 0.98279384
w1 = 0.2309
G = hnx.Hypergraph()
e0 = [("A", "B"), ("B", "C")]
e1 = [("A", "B", "C")]
G.add_edges_from(e0, weight=w0)
G.add_edges_from(e1, weight=w1)
	
print(G.adjacency_matrix())
print(G.adjacency_matrix(weight="weight"))

I have an issue with the edge adjacency matrix, because I would ordinarily calculate it with the following matrix formula: \sqrt(W) * M^T * M * \sqrt(W), but this won't currently work as edge_adjacency_matrix calls __incidence_to_adjacency, the same method as is called by adjacency_matrix. I would be grateful for your thoughts on this, I've currently left the weighted edge adjacency to throw a NotImplementedError.

All of these changes play quite nicely with the existing code, but I wanted to submit them for your consideration before doing too much more.

Many thanks. :)

Hi, this is a really good package :)

It would be loads more useful if you could provide support for weighted hypergraphs. It seems to be possible to create a weighted hypergraph at the moment like this:

import itertools
import hypernetx as hnx
import numpy as np

labels = "ABCD"
nodes = set(labels)
edges = list(
    itertools.chain(
        *[list(itertools.combinations(set(labels), ii)) for ii in range(2, len(labels))]
    )
)
hnx_edges = hnx.EntitySet(
    "Edges", 
    [hnx.Entity(
            "".join(ee), 
            ee,
            weight=np.random.rand() 
        ) for ee in edges]
)

h = hnx.Hypergraph(hnx_edges)
hnx.draw(h)
print(h.edges.elements)

but it's not possible to do any analysis with the weights as far as I can tell (see for example, the output of hnx.dist_stats(h) )

Are there any plans to add this type of functionality?

Thanks!

Installing hypernetx in a clean virtual environment fails due to a missing package: decorator.

Step to reproduce (after initializing a new empty virtual environment):

pip install hypernetx

giving an error like this:

>>> import hypernetx as hnx
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/__init__.py", line 7, in <module>
    from hypernetx.classes import *
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/classes/__init__.py", line 2, in <module>
    from .hypergraph import Hypergraph
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/classes/hypergraph.py", line 13, in <module>
    from hypernetx.classes.staticentity import StaticEntity, StaticEntitySet
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/classes/staticentity.py", line 10, in <module>
    from hypernetx.utils import (
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/utils/__init__.py", line 8, in <module>
    from .decorators import not_implemented_for
  File "/tmp/hnx-version/venv/lib/python3.8/site-packages/hypernetx/utils/decorators.py", line 4, in <module>
    from decorator import decorator
ModuleNotFoundError: No module named 'decorator'

Of course, installing the decorator package manually fixes the issue and hypernetx imports fine. I think the decorator package should just be added to the install_requires entry in setup.cfg and setup.py since this package is required to even load the package at all.

Tested using Python versions 3.8 and 3.9, using hypernetx 1.2.2 (on Ubuntu).

For pip install hypernetx, the packages igraph and celluloid are not installed by default. One has to do it manually. This is also the reason why the tutorial fails (#73). What about adding them to setup.py?

Hi, Maybe I am just blind (and/or stupid), but how can I identify (easily) all hyperedges of a given node n in a (static) hypergraph H in HyperNetX?

The solution I came up is:

[edge for edge, nodes in H.incidence_dict.items() if n in nodes]

But probably you had another, more performant solution in mind? Thank you!

Hello, I have the following error when I try to import hypernetx on Colab (with your tutorials) and also on my conda env:

ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-3-b7c66bb90caa> in <module>()
      3 import matplotlib.pyplot as plt
      4 import networkx as nx
----> 5 import hypernetx as hnx
      6 

4 frames
/usr/local/lib/python3.7/dist-packages/hypernetx/__init__.py in <module>()
      5 )
      6 from hypernetx.read_write import to_pickle, load_from_pickle
----> 7 from hypernetx.classes import *
      8 from hypernetx.reports import *
      9 from hypernetx.drawing import *

/usr/local/lib/python3.7/dist-packages/hypernetx/classes/__init__.py in <module>()
      1 from .entity import Entity, EntitySet
----> 2 from .hypergraph import Hypergraph
      3 from .staticentity import StaticEntity, StaticEntitySet

/usr/local/lib/python3.7/dist-packages/hypernetx/classes/hypergraph.py in <module>()
     11 from collections import OrderedDict, defaultdict
     12 from hypernetx.classes.entity import Entity, EntitySet
---> 13 from hypernetx.classes.staticentity import StaticEntity, StaticEntitySet
     14 from hypernetx.exception import HyperNetXError
     15 from hypernetx.utils.decorators import not_implemented_for

/usr/local/lib/python3.7/dist-packages/hypernetx/classes/staticentity.py in <module>()
      8 from hypernetx.exception import HyperNetXError
      9 from hypernetx.classes.entity import Entity, EntitySet
---> 10 from hypernetx.utils import HNXCount, DefaultOrderedDict, remove_row_duplicates
     11 from scipy.sparse import coo_matrix, csr_matrix, issparse
     12 import itertools as it

/usr/local/lib/python3.7/dist-packages/hypernetx/utils/__init__.py in <module>()
      2 from .decorators import not_implemented_for
      3 # from .toys import HarryPotter, LesMis, lesmis_hypergraph_from_df, book_tour, TransmissionProblem
----> 4 from .toys import *

ModuleNotFoundError: No module named 'hypernetx.utils.toys'

I'm doing something wrong?

When adding an existing node to an existing edge, the memberships property of the node stored in _nodes is not updated

Code to reproduce the issue:

system = hnx.Hypergraph()
edge0 = hnx.Entity('edge0')
edge1 = hnx.Entity('edge1')
node0 = hnx.Entity('node0')
node1 = hnx.Entity('node1')

system.add_edge(edge0)
system.add_edge(edge1)
system.add_node_to_edge(node0,edge0)
system.add_node_to_edge(node0,edge1)
system.add_node_to_edge(node1,edge1)

#Memberships seen from the nodes stored within _edges
print("Memberships seen from the nodes stored within _edges")
for el in system.edges:
    for edge in system.edges[el]:
        print(edge,system.edges[el][edge].memberships)
        
#Memberships seen from the nodes stored within _nodes
print("Memberships seen from the nodes stored within _nodes")
for el in system.nodes:
    print(el,system.nodes[el].memberships)

hnx.drawing.rubber_band.draw(system)

When drawing hypergraphs using the rubber band layout, node labels often overlap. This is especially prevalent when using longer strings as node labels; however, this problem also often occurs with very short labels as well. Due to the overlap, many labels are unreadable. Are there any settings within HyperNetX or Matplotlib to correct this? If not, are there any known workarounds? Thanks!

I made the following diagram using Hypernetworkx's draw command.

The node fst_score node appears really close to the 233 hyperedge, making it hard to distinguish whether the node is a member of the hyperedge. Is there a setting an argument somewhere that would allow me to ensure that the hyperedges maintain a minimal distance from the nodes?

Is it possible for edges with the same name but different nodes to coexist. Something like this :

{
  "is" : ["red","apple","rotten"],
  "is" : ["black","pen"]
}

How do you achieve this ?

Hello, thank you for sharing this excellent library.

I want to compare the performance of clustering on weighted and unweighted hypergraphs as described in K. Hayashi, S. Aksoy, C. Park, H. Park, "Hypergraph random walks, Laplacians, and clustering". I tried to do this based on Tutorial 11. However, I think some modifications need. h = hnx.Hypergraph(hnx.StaticEntitySet(data=data, weights=w)) I think this code does not make a weighted hypergraph. Instead, this should be revised as below. h = hnx.Hypergraph(hnx.StaticEntitySet(data=data, weights=w), weights=w)

Additionally, I cannot get the satisfying result that weighted hypergraphs perform better than unweighted ones. Thus, I want to ask whether the below code is a correct way to make clustering on an unweighted hypergraph and evaluate clustering algorithms by NMI scores.

# get data
categories = all_categories[[1,15]]
twenty_train = fetch_20newsgroups(subset='test', categories=categories, shuffle=True, random_state=42)
doc_types=dict()
for i,x in enumerate(twenty_train.filenames):
    doc_types[i]=x.split('/')[-2]
tfidf_vect = TfidfVectorizer()
X_tfidf = tfidf_vect.fit_transform(twenty_train.data)

# construct hypergraph
mat = coo_matrix(X_tfidf)
edges = mat.col
nodes = mat.row
data = np.array([edges,nodes]).T
weights = mat.data
# clustering on weighted hypergraph
h = hnx.Hypergraph(hnx.StaticEntitySet(data=data,weights=weights),weights=weights)
clusters=hnx.spec_clus(h,num_clus,weights=True)
# clustering on unweighted hypergraph
uwh =  hnx.Hypergraph(hnx.StaticEntitySet(data=data,weights=None))
uw_clusters = hnx.spec_clus(uwh,num_clus,weights=False)

# construct clustering label list
categoryindexing = {}
_labels = {} # answer
for i in range(X_tfidf.shape[0]):
    if doc_types[i] not in categoryindexing:
        categoryindexing[doc_types[i]] = len(categoryindexing)
    ci = categoryindexing[doc_types[i]]
    _labels[i] = ci
labels = [_labels[i] for i in range(X_tfidf.shape[0])]
_w_pred = {} #  labels from weighted hypergraph
for i in clusters:
    for v in clusters[i]:
        _w_pred[v] = i
w_pred = [_w_pred[i] for i in range(X_tfidf.shape[0])]
_uw_pred = {} # labels from unweighted hypergraph
for i in uw_clusters:
    for v in uw_clusters[i]:
        _uw_pred[v] = i
uw_pred = [_uw_pred[i] for i in range(X_tfidf.shape[0])]

# evaluation on NMI score
from sklearn.metrics.cluster import normalized_mutual_info_score as NMI_SCORE
score = NMI_SCORE(labels, w_pred)
print(score) # 0.73
score = NMI_SCORE(labels, uw_pred)
print(score) # 0.78

When I call print('qH =',hmod.modularity(H, K, strict)) it returns AttributeError: 'Entity' object has no attribute 'strength'

The full code:

import igraph as ig
import hypernetx as hnx
import hypernetx.algorithms.hypergraph_modularity as hmod

H = hnx.Hypergraph(lookup)
H = hmod.precompute_attributes(H)
#%%
K = hmod.kumar(H)
#%%
strict = hmod.strict
## Compute qH
print('qH =',hmod.modularity(H, K, strict))

Lookup is a dictionary of the form: {key1 : (hedge1, hedge2...), key2 : (hedge1, hedge3...)}

I tried to generate a hypergraph using node neighborhood information from a matrix K of size |N| x |N|. Using the simple graph edges list, if the length (distance) of two given nodes is less than or equal to k, then I put 1 in the K matrix, 0 otherwise. 1 denotes that the two given nodes are neighbors, and 0 denotes that nodes are not neighbors. After generating the K matrix, I used it to represent hyperedges where each row of K represents the neighborhood of a particular node. (a hyperedge).

The problem is, after generating a hypergraph, it shows the correct number of hyperedges using the H.edges method, however, it gives the wrong number of nodes using the H.nodes method. The total nodes are very less than the total nodes given by the H.nodes method,

Also, I tried the example graph given in the documentation of the library. It has 13 nodes and 8 hyperedges. H.nodes and H.edges methods show the correct number of nodes and edges for this graph.

Can't I use the defined K matrix for hypergraph generation?