Within connectionism, what is a back-propagation model? Give an example.
Connectionism was proposed by Donald Hebb in 1940s and is used in the studies of artificial intelligence. It is a set of approach, also utilized in other disciplines like cognitive psychology, neuroscience, and psychology. It models behavioral phenomena by interconnecting networks of simple units.
Backpropagation, also called backward propagation of errors, is a method of setting artificial neural networks in conjunction with an appropriate and optimized method. It calculates the gradient loss with respect to the network setup. The gradient is then further fed to an optimization method, which updates the data to minimize the loss of function.
In backpropagation, words are fed to the system, and the signals are passed through hidden units to phonemes. The phoneme then synthesizes the original voice and speech. This is just the similar replica of the model voice. This is how a Nettalk is possible.