What is the difference between feed forward neural network and non-linear autoregressive neural network. Do they have same structure. What is the difference in their equation
From what I understand—and please someone correct me if I am wrong—FFNN is less of a strict definition of a network and more a category that multiple networks are part of. In that sense, a NAR is an FFNN, but not necessarily the other way around. This is another "all squares are rectangles, but not are rectangles are squares" sort of thing.
But NAR is also not a single model, but rather a model type. This means that neither FFNNs nor NARs have a single equation that you could compare against one another.
Broadly, FFNNs have one-way data flows. This separates them from RNNs and CNNs that have loops built-in. NARs also consist of one-way data flows and are built on the math of autoregressive methods.