Sequential Estimation of Feed Forward Networks for Small Embedded Hardware

Abstract

Deterministic Finite Automata (DFA) and Non-Deterministic Automata (NFA) Systems require an input and lead to an output after stage traversals and based on the pathway so chosen leads to the either the acceptance or rejection of a language. Considering compilers, the compiler works first to understand the lexical correctness of the input and to do so follows steps to check for the validity of the same. If the input is of a valid form then the input is accepted else a suitable corresponding error is thrown. When considering a feed forward neural network, we see a pattern of an input being taken and passed to a hidden layer, which further may either pass to another hidden layer (making it a deep network) or lead it to an output layer. Neural networks find application in classification problems, regression analysis and recognition paradigms. On naïve speculation, a correlation can be made on similarities between finite automata and feed forward networks.