June 2020 - graph-tool - archives.skewed.de

Multiple edge weights and model comparison in the SBM
by James Ruffle 15 Jun '20

15 Jun '20

Dear Tiago / Graph Tool community, I am running the SBM on a directed network where edges are all conditional probability of one event, given the other, approximately 60 nodes. This is generated originally from a binary matrix of repeated samples and if these 60 nodes as variables occurred or did not. However, in this binary matrix I have some missing data, which cannot be easily imputed, and I suspect is not missing at random. I am keen to find a way to either control for this as a covariate, or alternatively show that the missingness of data does not cause a big issue in the SBM fit. One possibility I have been exploring is by computing a conditional probability network of the missingness, and passing that as an additional edge covariate to the model. So, in this instance I end up with two SBMs: - 1 with a single edge weight, the conditional probability of the events occurring. The community structure of this is meaningful in its interpretation. - 1 with two edge weights, the conditional probability of the events occurring, and also the conditional probability of the events being missing from the dataset (missingness matrix). The community structure of this is not as meaningful - largely events that are missing together will just cluster together. (For instance, I have several nodes representing given decades of life which will all just cluster together, as if being 20 is unknown then if they are 30 is also unknown.) What is the best way to robustly compare the two SBM models and (hopefully) illustrate the model fit is better with the conditional probability of the event alone? Alternatively, do you have any suggestions on how I would account for the missingness in the data when constructing the model? Thank you for your time, James

2 10

efficent way to assign values to vector property map
by BleakHeart 14 Jun '20

14 Jun '20

Hi tiago, I was thinking if there is an efficient way to assing values to a vector property map. The only working method that I can use is the following: vprop = g.new_vertex_property("vector<double>") for i in range(G.num_vertices()): vprop[G.vertex(i)] = [1,2,3] When I try to use something like this: index = np.where(some condition on vertex properties)[0] vprop.a[index] = [1,2,3] It gives me the error: "TypeError: 'NoneType' object does not support item assignment". The error comes only with vector< > type properties. I am using this to assing colours to some vertices if a given condition is satified. I hope that you can help, because the first method is really slow on graphs with hundred of thousands vertices. -- Sent from: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/

3 3

Change point detection: Inferring the mesoscale structure of layered, edge-valued, and time-varying networks
by rezccy 11 Jun '20

11 Jun '20

Hi, In the paper "Inferring the mesoscale structure of layered, edge-valued, and time-varying networks", you described the procedure of finding change points in layered networks. I am wondering if this function is implemented in the graph-tool library? I can't seem to find this implementation in the documentation. Many thanks. Rex -- Sent from: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/

2 3

ImportError: libboost_python3-py36.so.1.65.1
by Snehal Shekatkar 06 Jun '20

06 Jun '20

Hi Tiago, I am on Ubuntu 19.04, and I upgraded to the graph-tool 2.32 (edited sources.list, update and apt-get install). Now when I try to import graph tool using "import graph_tool.all as gt", I get the following error: import graph_tool.all as gt File "/usr/lib/python3/dist-packages/graph_tool/__init__.py", line 106, in <module> dl_import("from . import libgraph_tool_core as libcore") File "/usr/lib/python3/dist-packages/graph_tool/dl_import.py", line 58, in dl_import exec(import_expr, local_dict, global_dict) File "<string>", line 1, in <module> ImportError: libboost_python3-py36.so.1.65.1: cannot open shared object file: No such file or directory Could you kindly suggest a solution? Thanks and regards, Snehal Shekatkar

2 3

Re: [graph-tool] Error while logging in/registering git.skewed.de
by Tiago de Paula Peixoto 05 Jun '20

05 Jun '20

Am 05.06.20 um 07:07 schrieb kicasta: > Hi Tiago, > > I tried to register to git.skewed.de using my email, in order to open an > issue as discussed in another thread, but never received the confirmation > email. Afterward, I tried to login via my github account, but since it has > the same email as my first attempt it didn´t let me; fair enough. However, I > then tried to login using my Bitbucket account that is linked to the > university mail and was also unsucessful. See screenshot. > > <http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com…> > > Am I doing something wrong or is there another process to register the site. Could you please try again? I have been doing maintenance on the server, which has just been moved to a new location. -- Tiago de Paula Peixoto <tiago(a)skewed.de>

2 3

Re: [graph-tool] Distribution while fitting weighted SBM
by Dominik Schlechtweg 04 Jun '20

04 Jun '20

Hi Tiago, thanks, this is already quite helpful. Short follow-up: Our final aim is to sample from the fitted model, also according to the inferred edge covariate distributions between groups. Is this somehow possible with graph-tool? (Up to now we were not able to do this.) Would it, in any case, be valid to retrieve the empirical distributions between each group from the fitted model and to fit a non-microcanonical version of the distributions (like binomial) to the covariates for each group combination, which could then be used to sample a weighted SBM e.g. with graspy? Or do you see a more direct way? Best, Dominik > On Jun 4, 2020, at 11:37, Tiago de Paula Peixoto <tiago(a)skewed.de <mailto:tiago@skewed.de>> wrote: > > Am 04.06.20 um 10:09 schrieb kicasta: > > Hi Tiago, > > I have a short question regarding your implementation of the weighted SBM > described here: > > https://arxiv.org/pdf/1708.01432.pdf > https://graph-tool.skewed.de/static/doc/demos/inference/inference.html#edge… > > Does your implementation fit one distribution to sample edges per block > combination or one global distribution? > > > One distribution per pair of groups (otherwise the covariates and the > group structure would be completely decoupled, which would be the same > as ignoring them altogether.) > > And how to retrieve the parameters > of the distribution from a fitted model? > > > The distributions are "microcanonical", i.e. the parameters are > quantities like the total sum of covariates, which is not allowed to > fluctuate. For example the "exponential" distribution for nonegative > covariates assumes that they are uniformly distributed among all > possibilities that have the exact same sum. This means that the > parameters need not be explicitly encoded. So, in this case, if you want > to extract the parameter of the distribution, you just get the sum of > covariates between any two pairs of groups. > > We were trying to fit a model similar to the SBM in graspy used for > simulation: > > https://graspy.neurodata.io/tutorials/simulations/sbm.html > > There you can choose one distribution per block combination for simulation. > I was figuring that this should also be possible when fitting the model. > > > Yes, this exactly what is described in the above paper, and what is > implemented in graph-tool. > > Best, > Tiago >

2 1

Distance map returned from shortest_distance function misses entries of certain vertices
by Mathias Versichele 04 Jun '20

04 Jun '20

Maybe someone can help me out with this problem I posted on StackOverFlow: https://stackoverflow.com/questions/61955566/distance-map-returned-from-sho… Basically, I discovered that a call to shortest_distance returns a distance map where certain node entries are missing (I need 349, I only get 328). Weird thing is that some of the missing nodes are intermediaries for other nodes that *are* present in the distance map. The link contains a MWE, which I tested on an older py2 setup, as well as an up-to-date Docker container. Results are the same. Would be very grateful for a solution, because I need to be certain about results correctness before I can proceed with this (very fast!) library. -- Sent from: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/

2 3

Distribution while fitting weighted SBM
by kicasta 04 Jun '20

04 Jun '20

Hi Tiago, I have a short question regarding your implementation of the weighted SBM described here: https://arxiv.org/pdf/1708.01432.pdf https://graph-tool.skewed.de/static/doc/demos/inference/inference.html#edge… Does your implementation fit one distribution to sample edges per block combination or one global distribution? And how to retrieve the parameters of the distribution from a fitted model? We were trying to fit a model similar to the SBM in graspy used for simulation: https://graspy.neurodata.io/tutorials/simulations/sbm.html There you can choose one distribution per block combination for simulation. I was figuring that this should also be possible when fitting the model. Thanks for your help in advance! -- Sent from: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/

2 1

Programatically determining blocks on visualization with hSBM
by timhannigan 04 Jun '20

04 Jun '20

Hi, I've been using graph-tool extensively over the past 6 months with hSBM_topic modeling (https://github.com/martingerlach/hSBM_Topicmodel; from the Gerlach, Peixoto & Altmann, 2018 paper). I have one small question that may dramatically improve the usability of this. The output showing the hierarchical community structure in the word-document network (chord diagram) is beautiful. However, is there any way to programmatically identify the blocks (blue squares), so that I can map these onto the word groups (topics) produced by the hSBM model? Thanks, -Tim Hannigan -- Sent from: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/

2 1

Reduce function on parallel edges, and more general groupby-reduce in graph-tool
by Carlo Nicolini 04 Jun '20

04 Jun '20

I have a graph in graph-tool with parallel edges. All parallel edges are assigned an `EdgePropertyMap` representing their weight with a double. I would like to return a graph where the edges are condensed into a single edge where a custom reduce operation on these weights is performed. For example: H = gt.Graph(directed=True) H.add_edge(0,1) # edge 0 H.add_edge(0,1) # edge 1 H.add_edge(0,1) # edge 2 H.add_edge(1,0) # edge 3 H.add_edge(0,2) # edge 4 ew = H.new_edge_property("double") ew[list(H.edges())[0]]=1.2 ew[list(H.edges())[1]]=2.3 ew[list(H.edges())[2]]=-4.2 ew[list(H.edges())[3]]=5.8 ew[list(H.edges())[3]]=1.0 H.ep['weights'] = ew In this case the edge (0,1), with a reduce function "sum", should have total weight 1.2+2.3-4.2= -0.7 while the remaining edges should keep the same weight (i.e. edge 3 should keep weight 5.8 and edge 4 should maintain weight 1.0). Using `gt.condensation_graph` seems to do the trick, but only sum operation can be performed. To do this we need to use `H.vertex_index` as vertex property: Hcond, prop, vcount, ecount, v_avg, edge_avg = gt.condensation_graph(H,prop=H.vertex_index, aeprops=[H.ep['weights']]) Unfortunately this function only allows sum operation to be performed. What about using a different "reduce" function, like for example max? What about operations on non-numeric types of edges? It would greatly help carrying complex operations directly in Graph-tool, even if the value type of edges property is different than numeric. More generally, what about introducing a split-apply-reduce framework to vertices and edges, or is it already possible with some modifications to `gt.condensation_graph`?

2 1