• (choosen) https://networkx.github.io/documentation/networkx-1.10/reference/algorithms.flow.html
• https://developers.google.com/optimization/flow/maxflow
• https://graph-tool.skewed.de/static/doc/flow.html
• https://github.com/pmneila/PyMaxflow
• http://igraph.org/redirect.html

It is possible to solve this as an integer linear programming problem. That may be a more general approach.

sudo apt install python-pydot python-pydot-ng graphviz libgraphviz-dev
pip install -r requirements.txt

• invent better name

• design user interface
• design server architecture
• put it all together

constraints:

• randomize (? - to have multiple possible sets)

• indiference

• priority (co by učili více/méně rádi)

• multiple rounds v kt. stejní lidi neučí ve všech.

• check: learning set != teaching set

• multiple rounds

• in multiple rounds teach/learn one thing only once.

• no free people

  • automatic low-priority (cap 0.001) edges everybody -> learning everything

• nejsou skupiny o jednom člověku.

  • automatic.

MAYBE (future version):

• same topic taught in multiple groups at once