Richard Rubalcava - 1 year ago 154
Python Question

Using bfs coloring algorithm to check if graph is bipartite in Python

Code:

``````def bipartite(G):
open_list = [1]
colors = {}
color_counter = 0
# assign a color to the first node being visited
colors[1] = 0

while open_list:
# up the counter here so that all neighbors get the same color
color_counter += 1
# use first elem for bfs
current_neighbors = G[open_list[0]]
current_color = color_counter % 2
# prints used for debugging
print open_list
print "The current color is: %s" % (current_color,)
for neighbor in current_neighbors:
if neighbor not in colors:
open_list.append(neighbor)
colors[neighbor] = current_color
# print used for debugging
print "parent is: %s, child is: %s, %s's color is: %s" \
% (open_list[0], neighbor, neighbor, colors[neighbor])
# print used for debugging
else: print "parent is: %s, child is: %s, already colored: %s" \
% (open_list[0], neighbor, colors[neighbor])
open_list.pop(0)
# now, return array of values that has one of the two colors
zeros_array = []
ones_array = []
for key in colors.keys():
if colors[key] == 0:
zeros_array.append(key)
else:
ones_array.append(key)

if len(set(zeros_array) & set(ones_array)) == 0:
return zeros_array
else:
return None
``````

Here's the graph I'm using:

``````{1: {2: 1, 4: 1}, 2: {1: 1, 3: 1, 5: 1}, 3: {8: 1, 2: 1}, 4: {1: 1}, 5: {2: 1, 6: 1}, 6: {5: 1}, 8: {3: 1}}
``````

I drew it out and the graph can be visualized as a tree with 1 as the root, and branches off to nodes 2 and 4, where 4 is a leaf, but 2 keeps going. I'm using a color counter to color neighbors the same color (either 0 or 1). 2 and 4 are given the same color, then the algorithm correctly gives 3 and 5 the opposite color of their parent 2, but when returning one level up to check 4, the color counter is incremented, so by the time it gets to 8, 8 gets the wrong color.

I'm stuck at how to best fix this.

You should choose the color in depending on the your current vertex color, something like `colors[neighbor] = (colors[open_list[0]] + 1) % 2`
Also, `len(set(zeros_array) & set(ones_array)) == 0` will always be `true`, so you aren't checking is bipartite has gone well. You could check it in else branch of `if neighbor not in colors:`: just assert that your neighbour has different color with current vertex.