Neighboring nodes in the networkTest if directed graph is connectedWhy is NeighborhoodGraph so slow?How to add new nodes to an existing graph with fixed (coordinates) nodes?How do I upload a graph as an adjacency list and find the betweenness centrality?Arranging “ranked” nodes of a graph symmetricallyVertexLabels with Graph PropertiesNetwork with Radial Gradient Fill NodesHow to format vertices and control placement in a directed graphHow to label a large number of vertices using a list of namesColor the nodes according to certain valuesHighlight all paths in a graph below some threshold lengthSandbox algorithm for multifractal analysis of complex networks

Concept of linear mappings are confusing me

Why is this code 6.5x slower with optimizations enabled?

A newer friend of my brother's gave him a load of baseball cards that are supposedly extremely valuable. Is this a scam?

How is it possible for user's password to be changed after storage was encrypted? (on OS X, Android)

How to type dʒ symbol (IPA) on Mac?

Is there really no realistic way for a skeleton monster to move around without magic?

What defenses are there against being summoned by the Gate spell?

Can Medicine checks be used, with decent rolls, to completely mitigate the risk of death from ongoing damage?

Copycat chess is back

Why is an old chain unsafe?

Circuitry of TV splitters

Set-theoretical foundations of Mathematics with only bounded quantifiers

Is there a minimum number of transactions in a block?

How to make payment on the internet without leaving a money trail?

How did the USSR manage to innovate in an environment characterized by government censorship and high bureaucracy?

Copenhagen passport control - US citizen

Why is the design of haulage companies so “special”?

How does one intimidate enemies without having the capacity for violence?

What typically incentivizes a professor to change jobs to a lower ranking university?

How can bays and straits be determined in a procedurally generated map?

Why are only specific transaction types accepted into the mempool?

Could a US political party gain complete control over the government by removing checks & balances?

How to calculate implied correlation via observed market price (Margrabe option)

Infinite past with a beginning?



Neighboring nodes in the network


Test if directed graph is connectedWhy is NeighborhoodGraph so slow?How to add new nodes to an existing graph with fixed (coordinates) nodes?How do I upload a graph as an adjacency list and find the betweenness centrality?Arranging “ranked” nodes of a graph symmetricallyVertexLabels with Graph PropertiesNetwork with Radial Gradient Fill NodesHow to format vertices and control placement in a directed graphHow to label a large number of vertices using a list of namesColor the nodes according to certain valuesHighlight all paths in a graph below some threshold lengthSandbox algorithm for multifractal analysis of complex networks













5












$begingroup$


Consider the graph:



graph = 1 <-> 2, 1 <-> 4, 1 <-> 5, 1 <-> 8, 1 <-> 10, 1 <-> 26, 1 <-> 37, 1 <-> 42, 1 <-> 62, 1 <-> 86, 1 <-> 93, 1 <-> 100, 2 <-> 3, 2 <-> 7, 2 <-> 9, 2 <-> 12, 2 <-> 14, 2 <-> 17, 2 <-> 18, 2 <-> 25, 2 <-> 36, 2 <-> 41, 2 <-> 46, 2 <-> 50, 2 <-> 55, 2 <-> 72, 2 <-> 75, 3 <-> 6, 3 <-> 28, 3 <-> 34, 3 <-> 63, 4 <-> 13, 4 <-> 21, 5 <-> 20, 5 <-> 35, 5 <-> 40, 5 <-> 45, 5 <-> 48, 5 <-> 74, 6 <-> 31, 6 <-> 70, 9 <-> 11, 9 <-> 54, 9 <-> 67, 11 <-> 16, 11 <-> 24, 11 <-> 58, 11 <-> 60, 11 <-> 61, 11 <-> 65, 11 <-> 69, 12 <-> 27, 13 <-> 15, 13 <-> 33, 13 <-> 76, 14 <-> 30, 15 <-> 19, 15 <-> 96, 15 <-> 98, 16 <-> 57, 16 <-> 90, 19 <-> 22, 19 <-> 23, 19 <-> 39, 19 <-> 80, 19 <-> 83, 21 <-> 38, 22 <-> 59, 22 <-> 82, 25 <-> 29, 25 <-> 56, 25 <-> 94, 26 <-> 32, 26 <-> 43, 26 <-> 71, 27 <-> 47, 30 <-> 77, 30 <-> 78, 33 <-> 79, 33 <-> 97, 39 <-> 49, 39 <-> 51, 40 <-> 44, 40 <-> 73, 42 <-> 68, 48 <-> 52, 48 <-> 81, 50 <-> 53, 50 <-> 64, 50 <-> 89, 56 <-> 66, 56 <-> 92, 59 <-> 91, 62 <-> 88, 67 <-> 87, 74 <-> 95, 82 <-> 84, 82 <-> 85, 82 <-> 99;

net = Graph[graph, VertexShapeFunction -> "Name"]


Let's choose any node 'g' in the graph:



g=19;


Let 'r' denote the distance (counted in the number of nodes) from the node 'g':



d = GraphDiameter[net]
r = Range[1, d]


How to count all neighboring nodes within radius 'r' from the node 'g' ?



For example for node g=19 we have 6 nodes for r=1 (nodes: 80,83,22,39,23,15). For r=2 we have 7 nodes: 59,82,49,51,98,96,13.










share|improve this question











$endgroup$
















    5












    $begingroup$


    Consider the graph:



    graph = 1 <-> 2, 1 <-> 4, 1 <-> 5, 1 <-> 8, 1 <-> 10, 1 <-> 26, 1 <-> 37, 1 <-> 42, 1 <-> 62, 1 <-> 86, 1 <-> 93, 1 <-> 100, 2 <-> 3, 2 <-> 7, 2 <-> 9, 2 <-> 12, 2 <-> 14, 2 <-> 17, 2 <-> 18, 2 <-> 25, 2 <-> 36, 2 <-> 41, 2 <-> 46, 2 <-> 50, 2 <-> 55, 2 <-> 72, 2 <-> 75, 3 <-> 6, 3 <-> 28, 3 <-> 34, 3 <-> 63, 4 <-> 13, 4 <-> 21, 5 <-> 20, 5 <-> 35, 5 <-> 40, 5 <-> 45, 5 <-> 48, 5 <-> 74, 6 <-> 31, 6 <-> 70, 9 <-> 11, 9 <-> 54, 9 <-> 67, 11 <-> 16, 11 <-> 24, 11 <-> 58, 11 <-> 60, 11 <-> 61, 11 <-> 65, 11 <-> 69, 12 <-> 27, 13 <-> 15, 13 <-> 33, 13 <-> 76, 14 <-> 30, 15 <-> 19, 15 <-> 96, 15 <-> 98, 16 <-> 57, 16 <-> 90, 19 <-> 22, 19 <-> 23, 19 <-> 39, 19 <-> 80, 19 <-> 83, 21 <-> 38, 22 <-> 59, 22 <-> 82, 25 <-> 29, 25 <-> 56, 25 <-> 94, 26 <-> 32, 26 <-> 43, 26 <-> 71, 27 <-> 47, 30 <-> 77, 30 <-> 78, 33 <-> 79, 33 <-> 97, 39 <-> 49, 39 <-> 51, 40 <-> 44, 40 <-> 73, 42 <-> 68, 48 <-> 52, 48 <-> 81, 50 <-> 53, 50 <-> 64, 50 <-> 89, 56 <-> 66, 56 <-> 92, 59 <-> 91, 62 <-> 88, 67 <-> 87, 74 <-> 95, 82 <-> 84, 82 <-> 85, 82 <-> 99;

    net = Graph[graph, VertexShapeFunction -> "Name"]


    Let's choose any node 'g' in the graph:



    g=19;


    Let 'r' denote the distance (counted in the number of nodes) from the node 'g':



    d = GraphDiameter[net]
    r = Range[1, d]


    How to count all neighboring nodes within radius 'r' from the node 'g' ?



    For example for node g=19 we have 6 nodes for r=1 (nodes: 80,83,22,39,23,15). For r=2 we have 7 nodes: 59,82,49,51,98,96,13.










    share|improve this question











    $endgroup$














      5












      5








      5





      $begingroup$


      Consider the graph:



      graph = 1 <-> 2, 1 <-> 4, 1 <-> 5, 1 <-> 8, 1 <-> 10, 1 <-> 26, 1 <-> 37, 1 <-> 42, 1 <-> 62, 1 <-> 86, 1 <-> 93, 1 <-> 100, 2 <-> 3, 2 <-> 7, 2 <-> 9, 2 <-> 12, 2 <-> 14, 2 <-> 17, 2 <-> 18, 2 <-> 25, 2 <-> 36, 2 <-> 41, 2 <-> 46, 2 <-> 50, 2 <-> 55, 2 <-> 72, 2 <-> 75, 3 <-> 6, 3 <-> 28, 3 <-> 34, 3 <-> 63, 4 <-> 13, 4 <-> 21, 5 <-> 20, 5 <-> 35, 5 <-> 40, 5 <-> 45, 5 <-> 48, 5 <-> 74, 6 <-> 31, 6 <-> 70, 9 <-> 11, 9 <-> 54, 9 <-> 67, 11 <-> 16, 11 <-> 24, 11 <-> 58, 11 <-> 60, 11 <-> 61, 11 <-> 65, 11 <-> 69, 12 <-> 27, 13 <-> 15, 13 <-> 33, 13 <-> 76, 14 <-> 30, 15 <-> 19, 15 <-> 96, 15 <-> 98, 16 <-> 57, 16 <-> 90, 19 <-> 22, 19 <-> 23, 19 <-> 39, 19 <-> 80, 19 <-> 83, 21 <-> 38, 22 <-> 59, 22 <-> 82, 25 <-> 29, 25 <-> 56, 25 <-> 94, 26 <-> 32, 26 <-> 43, 26 <-> 71, 27 <-> 47, 30 <-> 77, 30 <-> 78, 33 <-> 79, 33 <-> 97, 39 <-> 49, 39 <-> 51, 40 <-> 44, 40 <-> 73, 42 <-> 68, 48 <-> 52, 48 <-> 81, 50 <-> 53, 50 <-> 64, 50 <-> 89, 56 <-> 66, 56 <-> 92, 59 <-> 91, 62 <-> 88, 67 <-> 87, 74 <-> 95, 82 <-> 84, 82 <-> 85, 82 <-> 99;

      net = Graph[graph, VertexShapeFunction -> "Name"]


      Let's choose any node 'g' in the graph:



      g=19;


      Let 'r' denote the distance (counted in the number of nodes) from the node 'g':



      d = GraphDiameter[net]
      r = Range[1, d]


      How to count all neighboring nodes within radius 'r' from the node 'g' ?



      For example for node g=19 we have 6 nodes for r=1 (nodes: 80,83,22,39,23,15). For r=2 we have 7 nodes: 59,82,49,51,98,96,13.










      share|improve this question











      $endgroup$




      Consider the graph:



      graph = 1 <-> 2, 1 <-> 4, 1 <-> 5, 1 <-> 8, 1 <-> 10, 1 <-> 26, 1 <-> 37, 1 <-> 42, 1 <-> 62, 1 <-> 86, 1 <-> 93, 1 <-> 100, 2 <-> 3, 2 <-> 7, 2 <-> 9, 2 <-> 12, 2 <-> 14, 2 <-> 17, 2 <-> 18, 2 <-> 25, 2 <-> 36, 2 <-> 41, 2 <-> 46, 2 <-> 50, 2 <-> 55, 2 <-> 72, 2 <-> 75, 3 <-> 6, 3 <-> 28, 3 <-> 34, 3 <-> 63, 4 <-> 13, 4 <-> 21, 5 <-> 20, 5 <-> 35, 5 <-> 40, 5 <-> 45, 5 <-> 48, 5 <-> 74, 6 <-> 31, 6 <-> 70, 9 <-> 11, 9 <-> 54, 9 <-> 67, 11 <-> 16, 11 <-> 24, 11 <-> 58, 11 <-> 60, 11 <-> 61, 11 <-> 65, 11 <-> 69, 12 <-> 27, 13 <-> 15, 13 <-> 33, 13 <-> 76, 14 <-> 30, 15 <-> 19, 15 <-> 96, 15 <-> 98, 16 <-> 57, 16 <-> 90, 19 <-> 22, 19 <-> 23, 19 <-> 39, 19 <-> 80, 19 <-> 83, 21 <-> 38, 22 <-> 59, 22 <-> 82, 25 <-> 29, 25 <-> 56, 25 <-> 94, 26 <-> 32, 26 <-> 43, 26 <-> 71, 27 <-> 47, 30 <-> 77, 30 <-> 78, 33 <-> 79, 33 <-> 97, 39 <-> 49, 39 <-> 51, 40 <-> 44, 40 <-> 73, 42 <-> 68, 48 <-> 52, 48 <-> 81, 50 <-> 53, 50 <-> 64, 50 <-> 89, 56 <-> 66, 56 <-> 92, 59 <-> 91, 62 <-> 88, 67 <-> 87, 74 <-> 95, 82 <-> 84, 82 <-> 85, 82 <-> 99;

      net = Graph[graph, VertexShapeFunction -> "Name"]


      Let's choose any node 'g' in the graph:



      g=19;


      Let 'r' denote the distance (counted in the number of nodes) from the node 'g':



      d = GraphDiameter[net]
      r = Range[1, d]


      How to count all neighboring nodes within radius 'r' from the node 'g' ?



      For example for node g=19 we have 6 nodes for r=1 (nodes: 80,83,22,39,23,15). For r=2 we have 7 nodes: 59,82,49,51,98,96,13.







      graphs-and-networks






      share|improve this question















      share|improve this question













      share|improve this question




      share|improve this question








      edited Apr 4 at 8:26









      J. M. is away

      98.9k10311467




      98.9k10311467










      asked Apr 4 at 8:25









      ralphralph

      1687




      1687




















          5 Answers
          5






          active

          oldest

          votes


















          5












          $begingroup$

          I will choose a bit better GraphLayout for a tree:



          net = Graph[graph, VertexLabels -> "Name", GraphLayout -> "RadialEmbedding"];


          I suggest don't just count directly - get an object - a subgraph - of your query, so you can then run various computations on it and don't need count all over again based on different criteria w/ a different code.



          nei[v_, d_] := NeighborhoodGraph[net, v, d]


          Take distance 1:



          nei[19, 1]


          enter image description here



          and see it is right:



          HighlightGraph[net, nei[19, 1]]


          enter image description here



          Now you can compute whatever you need:



          VertexList[nei[19, 1]]
          Length[%] - 1



          19, 15, 22, 23, 39, 80, 83



          6




          For the distance 2:



          VertexList[nei[19, 1]]
          VertexList[nei[19, 2]]
          Complement[%, %%]
          Length[%]



          19, 15, 22, 23, 39, 80, 83



          19, 13, 15, 22, 23, 39, 49, 51, 59, 80, 82, 83, 96, 98



          13, 49, 51, 59, 82, 96, 98



          7




          Timings for large graphs



          net = RandomGraph[BarabasiAlbertGraphDistribution[20000, 1]];

          nei[v_, d_] := NeighborhoodGraph[net, v, d]

          dist15:=Length[Complement[VertexList[nei[#,15]],VertexList[nei[#,14]]]&@RandomInteger[1000]]

          Table[AbsoluteTiming[dist15;][[1]], 5]



          0.097359, 0.094737, 0.092589, 0.08872, 0.087478







          share|improve this answer











          $endgroup$












          • $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
            $endgroup$
            – ralph
            Apr 4 at 12:24










          • $begingroup$
            @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 12:41










          • $begingroup$
            Please forgive me. I meant about 200,000 no 20,000 nodes.
            $endgroup$
            – ralph
            Apr 4 at 12:57










          • $begingroup$
            @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 13:51










          • $begingroup$
            @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01



















          3












          $begingroup$

          You could build it using BreadthFirstScan:



          net = RandomGraph[BarabasiAlbertGraphDistribution[200000, 1]];

          distance =
          GroupBy[Reap[
          BreadthFirstScan[net,
          19, "DiscoverVertex" -> (Sow[#3 -> #1] &)]][[2, 1]],
          First -> Last, Association["length" -> Length[#], "set" -> #] &];


          Get length:



          distance[3, "length"]



          1194




          distance[[All, "length"]]



          <|0 -> 1, 1 -> 214, 2 -> 1194, 3 -> 3058, 4 -> 5826, 5 -> 10069, 6
          -> 15110, 7 -> 19992, 8 -> 23821, 9 -> 24910, 10 -> 24767, 11 -> 21459, 12 -> 17869, 13 -> 13525, 14 -> 9119, 15 -> 5146, 16 -> 2406,
          17 -> 1025, 18 -> 337, 19 -> 106, 20 -> 34, 21 -> 11, 22 -> 1|>




          and set
          distance[21, "set"]




          182224, 145742, 171910, 124658, 125540, 128520, 196392, 166986,
          159530, 196846, 144772




          For weighted graphs:



          SeedRandom[123];net2 = Graph[net, EdgeWeight -> RandomInteger[1, 20, EdgeCount[net]]];

          edgeWeight[g_, x_, y_] :=
          With[weight = PropertyValue[g, UndirectedEdge[x, y],EdgeWeight],
          If[NumericQ[weight], weight, 0]]

          Clear[dist]; dist[_] := 0;
          weights =
          Reap[BreadthFirstScan[net2,
          9, "DiscoverVertex" -> ((dist[#1] =
          dist[#2] + edgeWeight[net2, #1, #2];
          Sow[#1 -> dist[#1]]) &)]][[2, 1]];

          set = Select[weights, #[[2]] <= 5 &];

          set[[;; 10]]



          9 -> 0, 66 -> 4, 126 -> 5, 160 -> 5, 190 -> 3, 274 -> 3, 283 -> 4,

          312 -> 4, 519 -> 5, 537 -> 4




          set // Length



          105




          Note that BreadthFirstScan approach might not work in general (non tree graphs).






          share|improve this answer











          $endgroup$








          • 1




            $begingroup$
            Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
            $endgroup$
            – Roman
            Apr 4 at 16:05






          • 1




            $begingroup$
            @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:14










          • $begingroup$
            @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:17











          • $begingroup$
            @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:20











          • $begingroup$
            @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
            $endgroup$
            – Szabolcs
            Apr 4 at 16:30


















          2












          $begingroup$

          To count how many nodes there are at every distance (unsorted Association): use this if you want to Lookup a particular distance:



          Counts@GraphDistance[net, g]



          <|4 -> 4, 5 -> 12, 3 -> 7, 6 -> 26, 7 -> 20, 2 -> 7, 8 -> 15, 1 -> 6, 0 -> 1, 9 -> 2|>




          Look them all up in order:



          BinCounts[GraphDistance[net, g], 0, d, 1]



          1, 6, 7, 7, 4, 12, 26, 20, 15, 2, 0, 0







          share|improve this answer











          $endgroup$












          • $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
            $endgroup$
            – ralph
            Apr 4 at 12:24










          • $begingroup$
            Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
            $endgroup$
            – Roman
            Apr 4 at 13:50


















          2












          $begingroup$


          How to count all neighboring nodes within radius 'r' from the node 'g' ?




          Use IGraph/M.



          IGNeighborhoodSize does precisely this and is probably your fastest bet, but I do not have time to benchmark it against other solutions right now.



          If you want to do it for multiple distances in one go, use IGDistanceCounts,



          IGDistanceCounts[graph, vertex]


          This gives you the counts of other vertices found at all (unweighted) distances. You can then simply Accumulate that list to get the result for all r at the same time.



          For weighted distances, use IGDistanceHistogram.






          share|improve this answer









          $endgroup$












          • $begingroup$
            Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
            $endgroup$
            – ralph
            Apr 4 at 14:15











          • $begingroup$
            @ralph As I said above, use IGDistanceHistogram
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01










          • $begingroup$
            Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
            $endgroup$
            – ralph
            2 days ago










          • $begingroup$
            @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
            $endgroup$
            – Szabolcs
            2 days ago










          • $begingroup$
            @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
            $endgroup$
            – Szabolcs
            2 days ago



















          2












          $begingroup$

          For weighted network:



          g1 = 4798 <-> 2641, 4798 <-> 2310, 4798 <-> 4721, 2310 <-> 1942,2310 <-> 961, 4721 <-> 4507, 4721 <-> 4779, 4779 <-> 4336, 4779 <-> 3238, 4336 <-> 3277, 4336 <-> 3514, 3277 <-> 2923, 2923 <-> 2772, 2923 <-> 2401, 2772 <-> 2, 2772 <-> 2771, 3514 <-> 3042, 3514 <-> 2739, 3042 <-> 3007, 3042 <-> 1655, 2739 <-> 2277, 2739 <-> 1895, 2 <-> 5, 2 <-> 3, 3277 <-> 100, 5 <-> 6, 5 <-> 7, 5 <-> 8, 5 <-> 9;

          w1 = 10, 20, 20, 4, 35, 3, 4, 6, 17, 7, 13, 2, 2, 7, 2, 1, 3, 5, 3, 6,4, 6, 2, 1, 1, 1, 1, 1, 1;

          w2=Table[1, 29];

          net1 = Graph[g1, EdgeWeight -> w1, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

          net2 = Graph[g1, EdgeWeight -> w2, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

          s = RandomSample[VertexList[net1], 15];

          Mr = Table[IGDistanceCounts[net1, s[[i]]], i, 1, Length[s]] (*for non weighted*)

          Mr2 = IGDistanceHistogram[net1, 9] (*for weighted graph ?*)

          Mr3 = IGDistanceHistogram[net2, 9] (*for non weighted graph ? Mr3==Mr *)





          share|improve this answer









          $endgroup$













            Your Answer





            StackExchange.ifUsing("editor", function ()
            return StackExchange.using("mathjaxEditing", function ()
            StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix)
            StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
            );
            );
            , "mathjax-editing");

            StackExchange.ready(function()
            var channelOptions =
            tags: "".split(" "),
            id: "387"
            ;
            initTagRenderer("".split(" "), "".split(" "), channelOptions);

            StackExchange.using("externalEditor", function()
            // Have to fire editor after snippets, if snippets enabled
            if (StackExchange.settings.snippets.snippetsEnabled)
            StackExchange.using("snippets", function()
            createEditor();
            );

            else
            createEditor();

            );

            function createEditor()
            StackExchange.prepareEditor(
            heartbeatType: 'answer',
            autoActivateHeartbeat: false,
            convertImagesToLinks: false,
            noModals: true,
            showLowRepImageUploadWarning: true,
            reputationToPostImages: null,
            bindNavPrevention: true,
            postfix: "",
            imageUploader:
            brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
            contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
            allowUrls: true
            ,
            onDemand: true,
            discardSelector: ".discard-answer"
            ,immediatelyShowMarkdownHelp:true
            );



            );













            draft saved

            draft discarded


















            StackExchange.ready(
            function ()
            StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmathematica.stackexchange.com%2fquestions%2f194581%2fneighboring-nodes-in-the-network%23new-answer', 'question_page');

            );

            Post as a guest















            Required, but never shown

























            5 Answers
            5






            active

            oldest

            votes








            5 Answers
            5






            active

            oldest

            votes









            active

            oldest

            votes






            active

            oldest

            votes









            5












            $begingroup$

            I will choose a bit better GraphLayout for a tree:



            net = Graph[graph, VertexLabels -> "Name", GraphLayout -> "RadialEmbedding"];


            I suggest don't just count directly - get an object - a subgraph - of your query, so you can then run various computations on it and don't need count all over again based on different criteria w/ a different code.



            nei[v_, d_] := NeighborhoodGraph[net, v, d]


            Take distance 1:



            nei[19, 1]


            enter image description here



            and see it is right:



            HighlightGraph[net, nei[19, 1]]


            enter image description here



            Now you can compute whatever you need:



            VertexList[nei[19, 1]]
            Length[%] - 1



            19, 15, 22, 23, 39, 80, 83



            6




            For the distance 2:



            VertexList[nei[19, 1]]
            VertexList[nei[19, 2]]
            Complement[%, %%]
            Length[%]



            19, 15, 22, 23, 39, 80, 83



            19, 13, 15, 22, 23, 39, 49, 51, 59, 80, 82, 83, 96, 98



            13, 49, 51, 59, 82, 96, 98



            7




            Timings for large graphs



            net = RandomGraph[BarabasiAlbertGraphDistribution[20000, 1]];

            nei[v_, d_] := NeighborhoodGraph[net, v, d]

            dist15:=Length[Complement[VertexList[nei[#,15]],VertexList[nei[#,14]]]&@RandomInteger[1000]]

            Table[AbsoluteTiming[dist15;][[1]], 5]



            0.097359, 0.094737, 0.092589, 0.08872, 0.087478







            share|improve this answer











            $endgroup$












            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 12:41










            • $begingroup$
              Please forgive me. I meant about 200,000 no 20,000 nodes.
              $endgroup$
              – ralph
              Apr 4 at 12:57










            • $begingroup$
              @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 13:51










            • $begingroup$
              @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01
















            5












            $begingroup$

            I will choose a bit better GraphLayout for a tree:



            net = Graph[graph, VertexLabels -> "Name", GraphLayout -> "RadialEmbedding"];


            I suggest don't just count directly - get an object - a subgraph - of your query, so you can then run various computations on it and don't need count all over again based on different criteria w/ a different code.



            nei[v_, d_] := NeighborhoodGraph[net, v, d]


            Take distance 1:



            nei[19, 1]


            enter image description here



            and see it is right:



            HighlightGraph[net, nei[19, 1]]


            enter image description here



            Now you can compute whatever you need:



            VertexList[nei[19, 1]]
            Length[%] - 1



            19, 15, 22, 23, 39, 80, 83



            6




            For the distance 2:



            VertexList[nei[19, 1]]
            VertexList[nei[19, 2]]
            Complement[%, %%]
            Length[%]



            19, 15, 22, 23, 39, 80, 83



            19, 13, 15, 22, 23, 39, 49, 51, 59, 80, 82, 83, 96, 98



            13, 49, 51, 59, 82, 96, 98



            7




            Timings for large graphs



            net = RandomGraph[BarabasiAlbertGraphDistribution[20000, 1]];

            nei[v_, d_] := NeighborhoodGraph[net, v, d]

            dist15:=Length[Complement[VertexList[nei[#,15]],VertexList[nei[#,14]]]&@RandomInteger[1000]]

            Table[AbsoluteTiming[dist15;][[1]], 5]



            0.097359, 0.094737, 0.092589, 0.08872, 0.087478







            share|improve this answer











            $endgroup$












            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 12:41










            • $begingroup$
              Please forgive me. I meant about 200,000 no 20,000 nodes.
              $endgroup$
              – ralph
              Apr 4 at 12:57










            • $begingroup$
              @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 13:51










            • $begingroup$
              @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01














            5












            5








            5





            $begingroup$

            I will choose a bit better GraphLayout for a tree:



            net = Graph[graph, VertexLabels -> "Name", GraphLayout -> "RadialEmbedding"];


            I suggest don't just count directly - get an object - a subgraph - of your query, so you can then run various computations on it and don't need count all over again based on different criteria w/ a different code.



            nei[v_, d_] := NeighborhoodGraph[net, v, d]


            Take distance 1:



            nei[19, 1]


            enter image description here



            and see it is right:



            HighlightGraph[net, nei[19, 1]]


            enter image description here



            Now you can compute whatever you need:



            VertexList[nei[19, 1]]
            Length[%] - 1



            19, 15, 22, 23, 39, 80, 83



            6




            For the distance 2:



            VertexList[nei[19, 1]]
            VertexList[nei[19, 2]]
            Complement[%, %%]
            Length[%]



            19, 15, 22, 23, 39, 80, 83



            19, 13, 15, 22, 23, 39, 49, 51, 59, 80, 82, 83, 96, 98



            13, 49, 51, 59, 82, 96, 98



            7




            Timings for large graphs



            net = RandomGraph[BarabasiAlbertGraphDistribution[20000, 1]];

            nei[v_, d_] := NeighborhoodGraph[net, v, d]

            dist15:=Length[Complement[VertexList[nei[#,15]],VertexList[nei[#,14]]]&@RandomInteger[1000]]

            Table[AbsoluteTiming[dist15;][[1]], 5]



            0.097359, 0.094737, 0.092589, 0.08872, 0.087478







            share|improve this answer











            $endgroup$



            I will choose a bit better GraphLayout for a tree:



            net = Graph[graph, VertexLabels -> "Name", GraphLayout -> "RadialEmbedding"];


            I suggest don't just count directly - get an object - a subgraph - of your query, so you can then run various computations on it and don't need count all over again based on different criteria w/ a different code.



            nei[v_, d_] := NeighborhoodGraph[net, v, d]


            Take distance 1:



            nei[19, 1]


            enter image description here



            and see it is right:



            HighlightGraph[net, nei[19, 1]]


            enter image description here



            Now you can compute whatever you need:



            VertexList[nei[19, 1]]
            Length[%] - 1



            19, 15, 22, 23, 39, 80, 83



            6




            For the distance 2:



            VertexList[nei[19, 1]]
            VertexList[nei[19, 2]]
            Complement[%, %%]
            Length[%]



            19, 15, 22, 23, 39, 80, 83



            19, 13, 15, 22, 23, 39, 49, 51, 59, 80, 82, 83, 96, 98



            13, 49, 51, 59, 82, 96, 98



            7




            Timings for large graphs



            net = RandomGraph[BarabasiAlbertGraphDistribution[20000, 1]];

            nei[v_, d_] := NeighborhoodGraph[net, v, d]

            dist15:=Length[Complement[VertexList[nei[#,15]],VertexList[nei[#,14]]]&@RandomInteger[1000]]

            Table[AbsoluteTiming[dist15;][[1]], 5]



            0.097359, 0.094737, 0.092589, 0.08872, 0.087478








            share|improve this answer














            share|improve this answer



            share|improve this answer








            edited Apr 4 at 12:44

























            answered Apr 4 at 9:11









            Vitaliy KaurovVitaliy Kaurov

            57.6k6162282




            57.6k6162282











            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 12:41










            • $begingroup$
              Please forgive me. I meant about 200,000 no 20,000 nodes.
              $endgroup$
              – ralph
              Apr 4 at 12:57










            • $begingroup$
              @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 13:51










            • $begingroup$
              @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01

















            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 12:41










            • $begingroup$
              Please forgive me. I meant about 200,000 no 20,000 nodes.
              $endgroup$
              – ralph
              Apr 4 at 12:57










            • $begingroup$
              @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
              $endgroup$
              – Vitaliy Kaurov
              Apr 4 at 13:51










            • $begingroup$
              @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01
















            $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
            $endgroup$
            – ralph
            Apr 4 at 12:24




            $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15).
            $endgroup$
            – ralph
            Apr 4 at 12:24












            $begingroup$
            @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 12:41




            $begingroup$
            @ralph is 0.1 seconds is slow? What timings do you need? No criteria for timings is mentioned in your original post.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 12:41












            $begingroup$
            Please forgive me. I meant about 200,000 no 20,000 nodes.
            $endgroup$
            – ralph
            Apr 4 at 12:57




            $begingroup$
            Please forgive me. I meant about 200,000 no 20,000 nodes.
            $endgroup$
            – ralph
            Apr 4 at 12:57












            $begingroup$
            @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 13:51




            $begingroup$
            @Szabolcs i was just answering question without performance consideration as it was not asked in the OP, which had a tiny graph. I added benchmark after he made a comment, and then he changed his comment again.
            $endgroup$
            – Vitaliy Kaurov
            Apr 4 at 13:51












            $begingroup$
            @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01





            $begingroup$
            @VitaliyKaurov Sorry about the comments, I was wrong: this was actually fixed in 12.0. That is why I deleted them.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01












            3












            $begingroup$

            You could build it using BreadthFirstScan:



            net = RandomGraph[BarabasiAlbertGraphDistribution[200000, 1]];

            distance =
            GroupBy[Reap[
            BreadthFirstScan[net,
            19, "DiscoverVertex" -> (Sow[#3 -> #1] &)]][[2, 1]],
            First -> Last, Association["length" -> Length[#], "set" -> #] &];


            Get length:



            distance[3, "length"]



            1194




            distance[[All, "length"]]



            <|0 -> 1, 1 -> 214, 2 -> 1194, 3 -> 3058, 4 -> 5826, 5 -> 10069, 6
            -> 15110, 7 -> 19992, 8 -> 23821, 9 -> 24910, 10 -> 24767, 11 -> 21459, 12 -> 17869, 13 -> 13525, 14 -> 9119, 15 -> 5146, 16 -> 2406,
            17 -> 1025, 18 -> 337, 19 -> 106, 20 -> 34, 21 -> 11, 22 -> 1|>




            and set
            distance[21, "set"]




            182224, 145742, 171910, 124658, 125540, 128520, 196392, 166986,
            159530, 196846, 144772




            For weighted graphs:



            SeedRandom[123];net2 = Graph[net, EdgeWeight -> RandomInteger[1, 20, EdgeCount[net]]];

            edgeWeight[g_, x_, y_] :=
            With[weight = PropertyValue[g, UndirectedEdge[x, y],EdgeWeight],
            If[NumericQ[weight], weight, 0]]

            Clear[dist]; dist[_] := 0;
            weights =
            Reap[BreadthFirstScan[net2,
            9, "DiscoverVertex" -> ((dist[#1] =
            dist[#2] + edgeWeight[net2, #1, #2];
            Sow[#1 -> dist[#1]]) &)]][[2, 1]];

            set = Select[weights, #[[2]] <= 5 &];

            set[[;; 10]]



            9 -> 0, 66 -> 4, 126 -> 5, 160 -> 5, 190 -> 3, 274 -> 3, 283 -> 4,

            312 -> 4, 519 -> 5, 537 -> 4




            set // Length



            105




            Note that BreadthFirstScan approach might not work in general (non tree graphs).






            share|improve this answer











            $endgroup$








            • 1




              $begingroup$
              Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
              $endgroup$
              – Roman
              Apr 4 at 16:05






            • 1




              $begingroup$
              @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:14










            • $begingroup$
              @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:17











            • $begingroup$
              @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:20











            • $begingroup$
              @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
              $endgroup$
              – Szabolcs
              Apr 4 at 16:30















            3












            $begingroup$

            You could build it using BreadthFirstScan:



            net = RandomGraph[BarabasiAlbertGraphDistribution[200000, 1]];

            distance =
            GroupBy[Reap[
            BreadthFirstScan[net,
            19, "DiscoverVertex" -> (Sow[#3 -> #1] &)]][[2, 1]],
            First -> Last, Association["length" -> Length[#], "set" -> #] &];


            Get length:



            distance[3, "length"]



            1194




            distance[[All, "length"]]



            <|0 -> 1, 1 -> 214, 2 -> 1194, 3 -> 3058, 4 -> 5826, 5 -> 10069, 6
            -> 15110, 7 -> 19992, 8 -> 23821, 9 -> 24910, 10 -> 24767, 11 -> 21459, 12 -> 17869, 13 -> 13525, 14 -> 9119, 15 -> 5146, 16 -> 2406,
            17 -> 1025, 18 -> 337, 19 -> 106, 20 -> 34, 21 -> 11, 22 -> 1|>




            and set
            distance[21, "set"]




            182224, 145742, 171910, 124658, 125540, 128520, 196392, 166986,
            159530, 196846, 144772




            For weighted graphs:



            SeedRandom[123];net2 = Graph[net, EdgeWeight -> RandomInteger[1, 20, EdgeCount[net]]];

            edgeWeight[g_, x_, y_] :=
            With[weight = PropertyValue[g, UndirectedEdge[x, y],EdgeWeight],
            If[NumericQ[weight], weight, 0]]

            Clear[dist]; dist[_] := 0;
            weights =
            Reap[BreadthFirstScan[net2,
            9, "DiscoverVertex" -> ((dist[#1] =
            dist[#2] + edgeWeight[net2, #1, #2];
            Sow[#1 -> dist[#1]]) &)]][[2, 1]];

            set = Select[weights, #[[2]] <= 5 &];

            set[[;; 10]]



            9 -> 0, 66 -> 4, 126 -> 5, 160 -> 5, 190 -> 3, 274 -> 3, 283 -> 4,

            312 -> 4, 519 -> 5, 537 -> 4




            set // Length



            105




            Note that BreadthFirstScan approach might not work in general (non tree graphs).






            share|improve this answer











            $endgroup$








            • 1




              $begingroup$
              Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
              $endgroup$
              – Roman
              Apr 4 at 16:05






            • 1




              $begingroup$
              @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:14










            • $begingroup$
              @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:17











            • $begingroup$
              @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:20











            • $begingroup$
              @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
              $endgroup$
              – Szabolcs
              Apr 4 at 16:30













            3












            3








            3





            $begingroup$

            You could build it using BreadthFirstScan:



            net = RandomGraph[BarabasiAlbertGraphDistribution[200000, 1]];

            distance =
            GroupBy[Reap[
            BreadthFirstScan[net,
            19, "DiscoverVertex" -> (Sow[#3 -> #1] &)]][[2, 1]],
            First -> Last, Association["length" -> Length[#], "set" -> #] &];


            Get length:



            distance[3, "length"]



            1194




            distance[[All, "length"]]



            <|0 -> 1, 1 -> 214, 2 -> 1194, 3 -> 3058, 4 -> 5826, 5 -> 10069, 6
            -> 15110, 7 -> 19992, 8 -> 23821, 9 -> 24910, 10 -> 24767, 11 -> 21459, 12 -> 17869, 13 -> 13525, 14 -> 9119, 15 -> 5146, 16 -> 2406,
            17 -> 1025, 18 -> 337, 19 -> 106, 20 -> 34, 21 -> 11, 22 -> 1|>




            and set
            distance[21, "set"]




            182224, 145742, 171910, 124658, 125540, 128520, 196392, 166986,
            159530, 196846, 144772




            For weighted graphs:



            SeedRandom[123];net2 = Graph[net, EdgeWeight -> RandomInteger[1, 20, EdgeCount[net]]];

            edgeWeight[g_, x_, y_] :=
            With[weight = PropertyValue[g, UndirectedEdge[x, y],EdgeWeight],
            If[NumericQ[weight], weight, 0]]

            Clear[dist]; dist[_] := 0;
            weights =
            Reap[BreadthFirstScan[net2,
            9, "DiscoverVertex" -> ((dist[#1] =
            dist[#2] + edgeWeight[net2, #1, #2];
            Sow[#1 -> dist[#1]]) &)]][[2, 1]];

            set = Select[weights, #[[2]] <= 5 &];

            set[[;; 10]]



            9 -> 0, 66 -> 4, 126 -> 5, 160 -> 5, 190 -> 3, 274 -> 3, 283 -> 4,

            312 -> 4, 519 -> 5, 537 -> 4




            set // Length



            105




            Note that BreadthFirstScan approach might not work in general (non tree graphs).






            share|improve this answer











            $endgroup$



            You could build it using BreadthFirstScan:



            net = RandomGraph[BarabasiAlbertGraphDistribution[200000, 1]];

            distance =
            GroupBy[Reap[
            BreadthFirstScan[net,
            19, "DiscoverVertex" -> (Sow[#3 -> #1] &)]][[2, 1]],
            First -> Last, Association["length" -> Length[#], "set" -> #] &];


            Get length:



            distance[3, "length"]



            1194




            distance[[All, "length"]]



            <|0 -> 1, 1 -> 214, 2 -> 1194, 3 -> 3058, 4 -> 5826, 5 -> 10069, 6
            -> 15110, 7 -> 19992, 8 -> 23821, 9 -> 24910, 10 -> 24767, 11 -> 21459, 12 -> 17869, 13 -> 13525, 14 -> 9119, 15 -> 5146, 16 -> 2406,
            17 -> 1025, 18 -> 337, 19 -> 106, 20 -> 34, 21 -> 11, 22 -> 1|>




            and set
            distance[21, "set"]




            182224, 145742, 171910, 124658, 125540, 128520, 196392, 166986,
            159530, 196846, 144772




            For weighted graphs:



            SeedRandom[123];net2 = Graph[net, EdgeWeight -> RandomInteger[1, 20, EdgeCount[net]]];

            edgeWeight[g_, x_, y_] :=
            With[weight = PropertyValue[g, UndirectedEdge[x, y],EdgeWeight],
            If[NumericQ[weight], weight, 0]]

            Clear[dist]; dist[_] := 0;
            weights =
            Reap[BreadthFirstScan[net2,
            9, "DiscoverVertex" -> ((dist[#1] =
            dist[#2] + edgeWeight[net2, #1, #2];
            Sow[#1 -> dist[#1]]) &)]][[2, 1]];

            set = Select[weights, #[[2]] <= 5 &];

            set[[;; 10]]



            9 -> 0, 66 -> 4, 126 -> 5, 160 -> 5, 190 -> 3, 274 -> 3, 283 -> 4,

            312 -> 4, 519 -> 5, 537 -> 4




            set // Length



            105




            Note that BreadthFirstScan approach might not work in general (non tree graphs).







            share|improve this answer














            share|improve this answer



            share|improve this answer








            edited 2 days ago

























            answered Apr 4 at 14:29









            halmirhalmir

            10.6k2544




            10.6k2544







            • 1




              $begingroup$
              Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
              $endgroup$
              – Roman
              Apr 4 at 16:05






            • 1




              $begingroup$
              @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:14










            • $begingroup$
              @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:17











            • $begingroup$
              @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:20











            • $begingroup$
              @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
              $endgroup$
              – Szabolcs
              Apr 4 at 16:30












            • 1




              $begingroup$
              Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
              $endgroup$
              – Roman
              Apr 4 at 16:05






            • 1




              $begingroup$
              @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:14










            • $begingroup$
              @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:17











            • $begingroup$
              @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
              $endgroup$
              – Szabolcs
              Apr 4 at 16:20











            • $begingroup$
              @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
              $endgroup$
              – Szabolcs
              Apr 4 at 16:30







            1




            1




            $begingroup$
            Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
            $endgroup$
            – Roman
            Apr 4 at 16:05




            $begingroup$
            Amazingly fast, halmir! Any idea why this solution is so much faster than GraphDistance, which I would have thought works by BreadthFirstScan internally?
            $endgroup$
            – Roman
            Apr 4 at 16:05




            1




            1




            $begingroup$
            @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:14




            $begingroup$
            @Roman I had the conviction that GraphDistance compute the entire GraphDistanceMatrix even if you gave it only one vertex. I do not remember what led me to this conclusion though. I do remember that I put a lot of effort into this functionality area in IGraph/M as I could not use M's built-ins for large graphs.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:14












            $begingroup$
            @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:17





            $begingroup$
            @Roman A qucik test tells me that on a tree (which is being benchmarked here) the complexity of GraphDistance is quadratic in the graph size even when given just one vertex. That should not be so.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:17













            $begingroup$
            @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:20





            $begingroup$
            @halmir Can you tell us whether this is a bug and if it is fixable? The quadratic complexity looks like a bug.
            $endgroup$
            – Szabolcs
            Apr 4 at 16:20













            $begingroup$
            @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
            $endgroup$
            – Szabolcs
            Apr 4 at 16:30




            $begingroup$
            @Roman I strongly suspect that I may have reported this issue to Wolfram in the past. See e.g. this post I wrote 3 years ago, where I mention it: mathematica.stackexchange.com/a/109408/12
            $endgroup$
            – Szabolcs
            Apr 4 at 16:30











            2












            $begingroup$

            To count how many nodes there are at every distance (unsorted Association): use this if you want to Lookup a particular distance:



            Counts@GraphDistance[net, g]



            <|4 -> 4, 5 -> 12, 3 -> 7, 6 -> 26, 7 -> 20, 2 -> 7, 8 -> 15, 1 -> 6, 0 -> 1, 9 -> 2|>




            Look them all up in order:



            BinCounts[GraphDistance[net, g], 0, d, 1]



            1, 6, 7, 7, 4, 12, 26, 20, 15, 2, 0, 0







            share|improve this answer











            $endgroup$












            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
              $endgroup$
              – Roman
              Apr 4 at 13:50















            2












            $begingroup$

            To count how many nodes there are at every distance (unsorted Association): use this if you want to Lookup a particular distance:



            Counts@GraphDistance[net, g]



            <|4 -> 4, 5 -> 12, 3 -> 7, 6 -> 26, 7 -> 20, 2 -> 7, 8 -> 15, 1 -> 6, 0 -> 1, 9 -> 2|>




            Look them all up in order:



            BinCounts[GraphDistance[net, g], 0, d, 1]



            1, 6, 7, 7, 4, 12, 26, 20, 15, 2, 0, 0







            share|improve this answer











            $endgroup$












            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
              $endgroup$
              – Roman
              Apr 4 at 13:50













            2












            2








            2





            $begingroup$

            To count how many nodes there are at every distance (unsorted Association): use this if you want to Lookup a particular distance:



            Counts@GraphDistance[net, g]



            <|4 -> 4, 5 -> 12, 3 -> 7, 6 -> 26, 7 -> 20, 2 -> 7, 8 -> 15, 1 -> 6, 0 -> 1, 9 -> 2|>




            Look them all up in order:



            BinCounts[GraphDistance[net, g], 0, d, 1]



            1, 6, 7, 7, 4, 12, 26, 20, 15, 2, 0, 0







            share|improve this answer











            $endgroup$



            To count how many nodes there are at every distance (unsorted Association): use this if you want to Lookup a particular distance:



            Counts@GraphDistance[net, g]



            <|4 -> 4, 5 -> 12, 3 -> 7, 6 -> 26, 7 -> 20, 2 -> 7, 8 -> 15, 1 -> 6, 0 -> 1, 9 -> 2|>




            Look them all up in order:



            BinCounts[GraphDistance[net, g], 0, d, 1]



            1, 6, 7, 7, 4, 12, 26, 20, 15, 2, 0, 0








            share|improve this answer














            share|improve this answer



            share|improve this answer








            edited Apr 4 at 12:19

























            answered Apr 4 at 9:04









            RomanRoman

            4,53011127




            4,53011127











            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
              $endgroup$
              – Roman
              Apr 4 at 13:50
















            • $begingroup$
              Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
              $endgroup$
              – ralph
              Apr 4 at 12:24










            • $begingroup$
              Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
              $endgroup$
              – Roman
              Apr 4 at 13:50















            $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
            $endgroup$
            – ralph
            Apr 4 at 12:24




            $begingroup$
            Thank you. The code gives correct results but is memory-consuming for large networks (around 200,000 nodes: net = RandomGraph [BarabasiAlbertGraphDistribution [20,000, 1] and d = 1,2,3,4, ..., 15)
            $endgroup$
            – ralph
            Apr 4 at 12:24












            $begingroup$
            Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
            $endgroup$
            – Roman
            Apr 4 at 13:50




            $begingroup$
            Yes if you want only short distances then @szabolcs has better tools available. This GraphDistance solution is only good if you want the distances to all nodes in the graph.
            $endgroup$
            – Roman
            Apr 4 at 13:50











            2












            $begingroup$


            How to count all neighboring nodes within radius 'r' from the node 'g' ?




            Use IGraph/M.



            IGNeighborhoodSize does precisely this and is probably your fastest bet, but I do not have time to benchmark it against other solutions right now.



            If you want to do it for multiple distances in one go, use IGDistanceCounts,



            IGDistanceCounts[graph, vertex]


            This gives you the counts of other vertices found at all (unweighted) distances. You can then simply Accumulate that list to get the result for all r at the same time.



            For weighted distances, use IGDistanceHistogram.






            share|improve this answer









            $endgroup$












            • $begingroup$
              Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
              $endgroup$
              – ralph
              Apr 4 at 14:15











            • $begingroup$
              @ralph As I said above, use IGDistanceHistogram
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01










            • $begingroup$
              Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
              $endgroup$
              – ralph
              2 days ago










            • $begingroup$
              @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
              $endgroup$
              – Szabolcs
              2 days ago










            • $begingroup$
              @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
              $endgroup$
              – Szabolcs
              2 days ago
















            2












            $begingroup$


            How to count all neighboring nodes within radius 'r' from the node 'g' ?




            Use IGraph/M.



            IGNeighborhoodSize does precisely this and is probably your fastest bet, but I do not have time to benchmark it against other solutions right now.



            If you want to do it for multiple distances in one go, use IGDistanceCounts,



            IGDistanceCounts[graph, vertex]


            This gives you the counts of other vertices found at all (unweighted) distances. You can then simply Accumulate that list to get the result for all r at the same time.



            For weighted distances, use IGDistanceHistogram.






            share|improve this answer









            $endgroup$












            • $begingroup$
              Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
              $endgroup$
              – ralph
              Apr 4 at 14:15











            • $begingroup$
              @ralph As I said above, use IGDistanceHistogram
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01










            • $begingroup$
              Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
              $endgroup$
              – ralph
              2 days ago










            • $begingroup$
              @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
              $endgroup$
              – Szabolcs
              2 days ago










            • $begingroup$
              @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
              $endgroup$
              – Szabolcs
              2 days ago














            2












            2








            2





            $begingroup$


            How to count all neighboring nodes within radius 'r' from the node 'g' ?




            Use IGraph/M.



            IGNeighborhoodSize does precisely this and is probably your fastest bet, but I do not have time to benchmark it against other solutions right now.



            If you want to do it for multiple distances in one go, use IGDistanceCounts,



            IGDistanceCounts[graph, vertex]


            This gives you the counts of other vertices found at all (unweighted) distances. You can then simply Accumulate that list to get the result for all r at the same time.



            For weighted distances, use IGDistanceHistogram.






            share|improve this answer









            $endgroup$




            How to count all neighboring nodes within radius 'r' from the node 'g' ?




            Use IGraph/M.



            IGNeighborhoodSize does precisely this and is probably your fastest bet, but I do not have time to benchmark it against other solutions right now.



            If you want to do it for multiple distances in one go, use IGDistanceCounts,



            IGDistanceCounts[graph, vertex]


            This gives you the counts of other vertices found at all (unweighted) distances. You can then simply Accumulate that list to get the result for all r at the same time.



            For weighted distances, use IGDistanceHistogram.







            share|improve this answer












            share|improve this answer



            share|improve this answer










            answered Apr 4 at 13:40









            SzabolcsSzabolcs

            163k14448945




            163k14448945











            • $begingroup$
              Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
              $endgroup$
              – ralph
              Apr 4 at 14:15











            • $begingroup$
              @ralph As I said above, use IGDistanceHistogram
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01










            • $begingroup$
              Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
              $endgroup$
              – ralph
              2 days ago










            • $begingroup$
              @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
              $endgroup$
              – Szabolcs
              2 days ago










            • $begingroup$
              @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
              $endgroup$
              – Szabolcs
              2 days ago

















            • $begingroup$
              Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
              $endgroup$
              – ralph
              Apr 4 at 14:15











            • $begingroup$
              @ralph As I said above, use IGDistanceHistogram
              $endgroup$
              – Szabolcs
              Apr 4 at 16:01










            • $begingroup$
              Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
              $endgroup$
              – ralph
              2 days ago










            • $begingroup$
              @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
              $endgroup$
              – Szabolcs
              2 days ago










            • $begingroup$
              @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
              $endgroup$
              – Szabolcs
              2 days ago
















            $begingroup$
            Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
            $endgroup$
            – ralph
            Apr 4 at 14:15





            $begingroup$
            Thanks. And how to count the same as the 'IGDistanceCounts[graph, vertex]' formula but for weighted networks?
            $endgroup$
            – ralph
            Apr 4 at 14:15













            $begingroup$
            @ralph As I said above, use IGDistanceHistogram
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01




            $begingroup$
            @ralph As I said above, use IGDistanceHistogram
            $endgroup$
            – Szabolcs
            Apr 4 at 16:01












            $begingroup$
            Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
            $endgroup$
            – ralph
            2 days ago




            $begingroup$
            Mr=IGDistanceHistogram[net1, ??] (*for weighted graph *) ???
            $endgroup$
            – ralph
            2 days ago












            $begingroup$
            @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
            $endgroup$
            – Szabolcs
            2 days ago




            $begingroup$
            @ralph Did you check the documentation? If you checked the documentation and you found it to be unclear, you are very welcome to suggest improvements.
            $endgroup$
            – Szabolcs
            2 days ago












            $begingroup$
            @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
            $endgroup$
            – Szabolcs
            2 days ago





            $begingroup$
            @ralph The syntax is IGDistanceHistogram[graph, binSize, vertex] where binSize is the bin size used for constructing the distance histogram. You must put the vertex in a list as the syntax also accepts multiple vertices.
            $endgroup$
            – Szabolcs
            2 days ago












            2












            $begingroup$

            For weighted network:



            g1 = 4798 <-> 2641, 4798 <-> 2310, 4798 <-> 4721, 2310 <-> 1942,2310 <-> 961, 4721 <-> 4507, 4721 <-> 4779, 4779 <-> 4336, 4779 <-> 3238, 4336 <-> 3277, 4336 <-> 3514, 3277 <-> 2923, 2923 <-> 2772, 2923 <-> 2401, 2772 <-> 2, 2772 <-> 2771, 3514 <-> 3042, 3514 <-> 2739, 3042 <-> 3007, 3042 <-> 1655, 2739 <-> 2277, 2739 <-> 1895, 2 <-> 5, 2 <-> 3, 3277 <-> 100, 5 <-> 6, 5 <-> 7, 5 <-> 8, 5 <-> 9;

            w1 = 10, 20, 20, 4, 35, 3, 4, 6, 17, 7, 13, 2, 2, 7, 2, 1, 3, 5, 3, 6,4, 6, 2, 1, 1, 1, 1, 1, 1;

            w2=Table[1, 29];

            net1 = Graph[g1, EdgeWeight -> w1, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

            net2 = Graph[g1, EdgeWeight -> w2, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

            s = RandomSample[VertexList[net1], 15];

            Mr = Table[IGDistanceCounts[net1, s[[i]]], i, 1, Length[s]] (*for non weighted*)

            Mr2 = IGDistanceHistogram[net1, 9] (*for weighted graph ?*)

            Mr3 = IGDistanceHistogram[net2, 9] (*for non weighted graph ? Mr3==Mr *)





            share|improve this answer









            $endgroup$

















              2












              $begingroup$

              For weighted network:



              g1 = 4798 <-> 2641, 4798 <-> 2310, 4798 <-> 4721, 2310 <-> 1942,2310 <-> 961, 4721 <-> 4507, 4721 <-> 4779, 4779 <-> 4336, 4779 <-> 3238, 4336 <-> 3277, 4336 <-> 3514, 3277 <-> 2923, 2923 <-> 2772, 2923 <-> 2401, 2772 <-> 2, 2772 <-> 2771, 3514 <-> 3042, 3514 <-> 2739, 3042 <-> 3007, 3042 <-> 1655, 2739 <-> 2277, 2739 <-> 1895, 2 <-> 5, 2 <-> 3, 3277 <-> 100, 5 <-> 6, 5 <-> 7, 5 <-> 8, 5 <-> 9;

              w1 = 10, 20, 20, 4, 35, 3, 4, 6, 17, 7, 13, 2, 2, 7, 2, 1, 3, 5, 3, 6,4, 6, 2, 1, 1, 1, 1, 1, 1;

              w2=Table[1, 29];

              net1 = Graph[g1, EdgeWeight -> w1, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

              net2 = Graph[g1, EdgeWeight -> w2, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

              s = RandomSample[VertexList[net1], 15];

              Mr = Table[IGDistanceCounts[net1, s[[i]]], i, 1, Length[s]] (*for non weighted*)

              Mr2 = IGDistanceHistogram[net1, 9] (*for weighted graph ?*)

              Mr3 = IGDistanceHistogram[net2, 9] (*for non weighted graph ? Mr3==Mr *)





              share|improve this answer









              $endgroup$















                2












                2








                2





                $begingroup$

                For weighted network:



                g1 = 4798 <-> 2641, 4798 <-> 2310, 4798 <-> 4721, 2310 <-> 1942,2310 <-> 961, 4721 <-> 4507, 4721 <-> 4779, 4779 <-> 4336, 4779 <-> 3238, 4336 <-> 3277, 4336 <-> 3514, 3277 <-> 2923, 2923 <-> 2772, 2923 <-> 2401, 2772 <-> 2, 2772 <-> 2771, 3514 <-> 3042, 3514 <-> 2739, 3042 <-> 3007, 3042 <-> 1655, 2739 <-> 2277, 2739 <-> 1895, 2 <-> 5, 2 <-> 3, 3277 <-> 100, 5 <-> 6, 5 <-> 7, 5 <-> 8, 5 <-> 9;

                w1 = 10, 20, 20, 4, 35, 3, 4, 6, 17, 7, 13, 2, 2, 7, 2, 1, 3, 5, 3, 6,4, 6, 2, 1, 1, 1, 1, 1, 1;

                w2=Table[1, 29];

                net1 = Graph[g1, EdgeWeight -> w1, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

                net2 = Graph[g1, EdgeWeight -> w2, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

                s = RandomSample[VertexList[net1], 15];

                Mr = Table[IGDistanceCounts[net1, s[[i]]], i, 1, Length[s]] (*for non weighted*)

                Mr2 = IGDistanceHistogram[net1, 9] (*for weighted graph ?*)

                Mr3 = IGDistanceHistogram[net2, 9] (*for non weighted graph ? Mr3==Mr *)





                share|improve this answer









                $endgroup$



                For weighted network:



                g1 = 4798 <-> 2641, 4798 <-> 2310, 4798 <-> 4721, 2310 <-> 1942,2310 <-> 961, 4721 <-> 4507, 4721 <-> 4779, 4779 <-> 4336, 4779 <-> 3238, 4336 <-> 3277, 4336 <-> 3514, 3277 <-> 2923, 2923 <-> 2772, 2923 <-> 2401, 2772 <-> 2, 2772 <-> 2771, 3514 <-> 3042, 3514 <-> 2739, 3042 <-> 3007, 3042 <-> 1655, 2739 <-> 2277, 2739 <-> 1895, 2 <-> 5, 2 <-> 3, 3277 <-> 100, 5 <-> 6, 5 <-> 7, 5 <-> 8, 5 <-> 9;

                w1 = 10, 20, 20, 4, 35, 3, 4, 6, 17, 7, 13, 2, 2, 7, 2, 1, 3, 5, 3, 6,4, 6, 2, 1, 1, 1, 1, 1, 1;

                w2=Table[1, 29];

                net1 = Graph[g1, EdgeWeight -> w1, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

                net2 = Graph[g1, EdgeWeight -> w2, EdgeLabels -> "EdgeWeight", VertexShapeFunction -> "Name"]

                s = RandomSample[VertexList[net1], 15];

                Mr = Table[IGDistanceCounts[net1, s[[i]]], i, 1, Length[s]] (*for non weighted*)

                Mr2 = IGDistanceHistogram[net1, 9] (*for weighted graph ?*)

                Mr3 = IGDistanceHistogram[net2, 9] (*for non weighted graph ? Mr3==Mr *)






                share|improve this answer












                share|improve this answer



                share|improve this answer










                answered Apr 4 at 17:40









                ralphralph

                1687




                1687



























                    draft saved

                    draft discarded
















































                    Thanks for contributing an answer to Mathematica Stack Exchange!


                    • Please be sure to answer the question. Provide details and share your research!

                    But avoid


                    • Asking for help, clarification, or responding to other answers.

                    • Making statements based on opinion; back them up with references or personal experience.

                    Use MathJax to format equations. MathJax reference.


                    To learn more, see our tips on writing great answers.




                    draft saved


                    draft discarded














                    StackExchange.ready(
                    function ()
                    StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmathematica.stackexchange.com%2fquestions%2f194581%2fneighboring-nodes-in-the-network%23new-answer', 'question_page');

                    );

                    Post as a guest















                    Required, but never shown





















































                    Required, but never shown














                    Required, but never shown












                    Required, but never shown







                    Required, but never shown

































                    Required, but never shown














                    Required, but never shown












                    Required, but never shown







                    Required, but never shown







                    Popular posts from this blog

                    Sum ergo cogito? 1 nng

                    419 nièngy_Soadمي 19bal1.5o_g

                    Queiggey Chernihivv 9NnOo i Zw X QqKk LpB