by Michele Laurelli
The region of input space that affects a particular neuron's activation in a neural network.
In CNNs, receptive field grows with each layer. Deeper neurons see larger input regions. Determined by filter size, stride, and network depth. Critical for understanding what information neurons can access.
3x3 filter has 3x3 receptive field
Deeper layers see larger regions
Dilated convolutions increase receptive field