CNN architecture is feed forward whereas RNNs not only take the inputs given by the user, but it also is fed the nth output(s) as one of the inputs. This is an important distinction between the models usage. The CNN model is best used for image classification and other classification problems that do not involve a sequence of events. The RNN is best utilized for its ability to retain memory from past data and is used mostly in time series applications or series applications where the past is relevant to the next data point.