⚙️ Simulator: Artificial Neuron in a CNN
y = f(∑ wᵢxᵢ + b)
x₁
w₁
x₂
w₂
b
The values of x range from -3 to 3 because, in a CNN, pixels (0–255) are normalized before entering the network. The drawing on the side translates that normalized value back into a gray tone, just for visual intuition — the numbers that count for the calculation are the ones on the bars.
Receptive Field → Convolution → Feature Map
Activation at f(z)
x₁, x₂ = intensity of pixels in the receptive field · w₁, w₂ = kernel (filter) weights · z = convolution result at this position · y = pixel value produced in the feature map, after activation.
z = (1.00 × 0.80) + (-1.50 × 0.50) + 0.20 = 0.25 → y = 0.25
Description: Figure 9.3: Interactive simulator of the Artificial Neuron in a CNN context: toggle between a convolutional layer neuron (where x_i are pixel intensities in a receptive field and w_i are kernel weights) and a fully connected layer neuron, adjusting inputs, weights, bias, and activation function to visualize the computation of z and output y in real time.