You must log in to use subscribtions.

# Hinton diagrams with matplotlib

Hinton diagrams are a way of visualizing numerical values in a matrix/vector, popular in the neural networks and machine learning literature. The area occupied by a square is proportional to a value's magnitude, and the colour (black or white in this case) indicates its sign (positive/negative).

```   1 import numpy as N
2 import pylab as P
3
4 def _blob(x,y,area,colour):
5     """
6     Draws a square-shaped blob with the given area (< 1) at
7     the given coordinates.
8     """
9     hs = N.sqrt(area) / 2
10     xcorners = N.array([x - hs, x + hs, x + hs, x - hs])
11     ycorners = N.array([y - hs, y - hs, y + hs, y + hs])
12     P.fill(xcorners, ycorners, colour, edgecolor=colour)
13
14 def hinton(W, maxWeight=None):
15     """
16     Draws a Hinton diagram for visualizing a weight matrix.
17     Temporarily disables matplotlib interactive mode if it is on,
18     otherwise this takes forever.
19     """
20     reenable = False
21     if P.isinteractive():
22         P.ioff()
23     P.clf()
24     height, width = W.shape
25     if not maxWeight:
26         maxWeight = 2**N.ceil(N.log(N.max(N.abs(W)))/N.log(2))
27
28     P.fill(N.array([0,width,width,0]),N.array([0,0,height,height]),'gray')
29     P.axis('off')
30     P.axis('equal')
31     for x in xrange(width):
32         for y in xrange(height):
33             _x = x+1
34             _y = y+1
35             w = W[y,x]
36             if w > 0:
37                 _blob(_x - 0.5, height - _y + 0.5, min(1,w/maxWeight),'white')
38             elif w < 0:
39                 _blob(_x - 0.5, height - _y + 0.5, min(1,-w/maxWeight),'black')
40     if reenable:
41         P.ion()
42     P.show()
```

Cookbook/Matplotlib/HintonDiagrams (last edited 2010-03-30 18:58:39 by MartinSpacek)