/nrg_simRank

Primary LanguagePython

Revision modifications

Graound truth for the simrank

  • read d3plot displacements

  • one time step, last

    • calcualte the diff
    • all time steps
    • just 5 parts

  • Make the todo tasks

  • Compare 3001

    • Simrank matrix value
      • argsort issue with small value, numpy.array([1., 0.5, 1.]).argsort(kind="heapsort")
    • Simrank loop
    • make feature graph
    • Pe vs IE weight
      • why from single to all paired sim differ
      • Have two different weight
      • Correct kg01 loading

  • Similarity from RMSE

  • How is it with component graph

  • Diagnoal value not 1 > effect of spread

    • Look in paper
    • Move the test functions

  • Run simrank for each simulation and ref seprately

  • What does SimRank do

  • Change the spread function

Extended data debug

  • get top color
  • decrease number of simulation range

Write result

  • update displacements
  • update text
  • write extended result

SimRank issue

  • spread normalization differ from simrank weighted normalization. spread normalize vs source edges where simRank normalizes vs target edges

  • recalculate spread

  • why sim > 1

    • check sim for test_epread > not bigger than 1
    • check combining sprd and evd

  • change simRankpp function to simrank_pp_similarity_numpy in nrg_simRank/inv_simRnk_extnd.py

  • clean sprd, evd, C option in nrg_simRank/inv_simRnk_extnd.py

  • run the extnd result

  • check simrank result with github functions > wrong result on the simple example

  • transfer result to the plot_nrg_extnd

  • spread only for PID nodes

  • change weight manipulation to weight subtraction

Make new result

  • Pe plot with selecting only closest

update cevt result

  • run the result
  • change the method
  • verify the result
    • whey there is similarities close to zero, is it 0 or sth else> tehre are smaller than 0.11 but not zero
    • why hh-ll doen't pick small HH anymore > errList is active
    • check the KDE if the smlll similarities belong to pid than sims > the errList was empty

scatter plot comparison

  • the color code, L sim to be red

  • legend order

  • save four plot

  • add four pic in tex

  • legend inside

  • fix the bolt for plot_report.py

  • section 6.2, 4 pid -> components

  • split the paper

  • cite the conf paper modeling

Graph visualization

  • make same vis as paper
    • check the error list > it was ok, the issue was that initital plot was without weight - one des is missing in curre
  • show loop with removing high distance nodes
    • why network2tikzi doesn't plot > issue with latex, set the style with plotly
  • show visualization with more releases / several rls together

SimRank with grouping

  • add grouping to the database
  • load the features
  • simrank query to set graph for groping-des COG - make new design based on box grouping - add the weights to the edge of box design - add intiger name of the box to the graph