Strong positive weight
Strong negative weight
Near zero
High activation
0
Epoch
—
Train loss
—
Val accuracy
Click ▶ Train to start — the network will adjust its weights on every epoch
Loss train val
Accuracy train val
Confusion Matrix — rows = actual, cols = predicted
Waiting for training to begin…
How it works
Each wine has 11 measurements (acidity, sugar, alcohol…). The network multiplies each by a weight, sums them up, and passes the result through an activation function — repeating this through each hidden layer.
The line colors show weights: blue = positive (this input pushes toward higher quality), red = negative (it pulls away). Thickness = strength.
Node brightness shows average activation: brighter = that neuron fires more often. Watch the network find which chemicals really matter.