Topic List

Click a topic to learn more.


Term of the Moment

cellular generations


Look Up Another Term


Redirected from: CNN

Definition: convolutional neural network


A widely used AI neural network that recognizes grid patterns. Using image recognition as the primary example, a convolutional neural network (CNN) breaks the input into pixels and connects them to a series of neuron layers. The output layer is the best interpretation of the input. Contrast with recurrent neural network, which is a widely used AI architecture for speech and natural language recognition. See neural network, machine learning and AI.






Character Recognition
This simplistic example shows how the letter "Z" is recognized by assigning each pixel in the matrix a number from 0 to 1 based on the intensity of its shade or color. All 784 pixels of the 28x28 matrix become the input layer, and each layer is mathematically connected to the next layer. The output is the weighted average of the pixels. See AI weights and biases.






There Are Many Kinds of CNNs
Neural networks can be complex. These are examples of convolutional architectures from the Asimov Institute. (Images courtesy of Fjodor van Veen and Stefan Leijnen (2019). The Neural Network Zoo.)