Loading slide

Loading contents...

[██████████████░░░░][█████████████████████░░░░░░░]25 / 33
<back>next

Several boundaries can do it

By giving the job to more than one neuron.

One neuron marks the small, pale safe corner. A second marks the large, dark safe corner.

Their answers become inputs to a final neuron. If either earlier neuron found a safe region, the final answer is "safe."

We have seen this pattern before. A switch did one tiny thing. Connected switches became gates, and connected gates became circuits.

Neurons follow the same pattern. Each makes one simple division. Connected together, they can make a decision no neuron could make alone.

Connect neurons this way and the result is a neural network.

# citations(1)↓
  1. [1]deeplearningbook.org