etc.
•
The learning rule for a sigmoid belief
net is:
•
With replicated weights this becomes:
h
2
v
2
h
1
v
1
h
0
v
0
