Prerequisites
Nueron perceptron model - linear classification - loss function - training - Back Propogation.
Why do we need CNN
- The number of parmeters learned in regular neural network will grow if we simplye use a fully connected network.
- Instead if we use a convolution mask to learn the wieghts, then the convolutiona mask wieghts are shared despite the size of the image. This reduces the no of weights to be learned.
- Below is a complex layout of CNN architecture called AlexNet.
