Watch a neural network learn the XOR function step by step
① Select an Input
② Watch Data Flow Through the Network
Input
0
A
0
B
Hidden
?
H1
?
H2
Output
?
Result
③ The Math (Step by Step)
Select an input above to see the calculation
④ Train the Network
0
Training Steps
-
Total Error
How training works: Each step, we show the network all 4 XOR examples.
When it gets one wrong, we calculate which weights were responsible for the error
(backpropagation), then nudge them slightly in the right direction. After enough
steps, the weights converge to values that correctly compute XOR.