/Network_statistics

Repo for notebooks / project for the 2AMS30 Network Statistics for Data Science course.

Primary LanguageJupyter Notebook

Network_statistics

Repo for notebooks / project for the 2AMS30 Network Statistics for Data Science course.

Install dependencies :

cd deps
bash installer.sh
cd ..
source networkStats/bin/activate

Spectral Clustering:

First Presentation:

Options:

  • -h, --help show this help message and exit

  • --mode N Decide which is the right output for the CLI.

  • --verbose N Print what the script is doing in stdout.

  • --n_init N Number of iterations for k_means.

  • --save_log N Create a file containing all of the logs.

  • --plt N Plot the generated data

  • --n_vals N [N ...] Number of vertices for SBM.

  • --n _classes N Number of community classes for SBM.

  • --output_dir N Output directory for storing images and log file.

To run the config used to plot the graphs used in the Presentation type:

cd projects/spectral_clustering_pt1
python main.py --mode "all" --save_log True --plt True --verbose True --n_vals 10 100 1000

Second Presentation:

Options:

  • --h, --help show this help message and exit
  • --output_dir N Output directory for storing images and log file.
  • --plt N Plot the generated data
  • --save_log N Decide if it has to save the log file or not
  • --cc_analysis N do the analys for every single connected component
  • --eigen_gap N Index for the slicing during eigen gap process, if the value is set to -1 the user will be prompted to insert the input via stdin
  • --cluster_method N Wich clustering methods has to be used on eigen Vectors.
  • --n_class N Number of classes the codes tries to infer.b
  • --laplacian N Diffrent types of Laplacian Matrix
  • --pruning N If we prune the graph or not
  • --pruning_factor N Factor for the maximum size of the connected components that get pruned

To run the clustering algorithm on each connected component type with n>20 type:

cd projects/spectral_clustering_pt2
python main.py

To run the clustering algorithm on the whole pruned graph (still with n>20) type:

cd projects/spectral_clustering_pt2
python main.py --cc_analysis False

Link Prediction:

All the analysis we did for the second presentation on link prediction
using similarity methods can be found inside the following iPython notebook:

cd projects/link_predction
# You can open it by simply using Visual Studio Code:
code main.ipynb 
``