Linear neurons

• These are simple but computationally limited – If we can make them learn we may get insight into more complicated neurons

bias

th

y

0

weight on

0

output

th

input

index over input connections