test_F1_MMR.py

# -*- coding: utf-8 -*-
"""
Created on Thu Feb 11 17:24:13 2021

@author: Meghana
"""
from numpy import random as rnd
import numpy as np

import networkx as nx
# requires networkx 2.5 (2.3 does'nt support)



def f1_mmr(wt_mat):
    wt_mat = - wt_mat
    
    nr, nc = np.shape(wt_mat)
    
    B = nx.Graph()
    
    # Add nodes with the node attribute "bipartite"
    
    B.add_nodes_from(['t' + str(i) for i in range(nr)], bipartite=0)
    
    B.add_nodes_from(['p' + str(j) for j in range(nc)], bipartite=1)
    
    # Add edges only between nodes of opposite node sets
    
    B.add_weighted_edges_from([('t' + str(i),'p' + str(j), wt_mat[i,j]) for i in range(nr) for j in range(nc)])
    
    
    max_matching_edges = nx.algorithms.bipartite.matching.minimum_weight_full_matching(B)
    
    sum_wts = -sum([B[key][val]['weight'] for key,val in max_matching_edges.items()])/2
    
    recall = sum_wts/nr
    
    prec = sum_wts/nc
    
    f1 = 2*recall*prec/(prec+recall)
    
    return prec, recall, f1, max_matching_edges

# test
#wt_mat = rnd.rand(3,4)
#prec, recall, f1, max_matching_edges = f1_mmr(wt_mat)
#print(f1)
#print(max_matching_edges)
#print(wt_mat)