Pattern recognition visualised
How does a neural network actually recognise patterns?
A trained network answers a seemingly simple question: does the image show a dog?
The animation shows both the data flow through the entire network and, in detail, what happens inside a single artificial neuron: weighting, summation, activation and propagation of signals.
This makes visible how pattern recognition emerges step by step from many simple computations.
Animation: how a trained neural network answers whether an image shows a dog
Step 1 · 0%
data flow through the network · with detail view
Step 1 — Input signals reach Hidden 1
The input image — a cat — is fed into the network.
Sum Σ
0.00
Activation σ(Σ)
0.00
Answer
—