Neutrosophic Deep Neural Network and Its Application to Soil Microbe Analysis

Abstract

An overview of a Neutrosophic Deep Neural Network (NDNN) is given here to illustrate how artificial neural networks and neutrosophic sets can be combined to profit from the comprehensive truth, indeterminate, and falsity degrees known in neutrosophic logic. The investigation details how the NDNN architecture was gradually enriched from its initial and simplest structure, which was a Neutrosophic Neural Network (NNN) with only a one-layer neuron. Next, it progresses to an NNN with one layer and one multi-input neuron, and then becomes even more sophisticated as an NNN with one layer and several multi-input neurons. As a result, the framework concludes with a real NDNN made up of multiple layers, and every layer includes several multi-input neurons. For each network structure, new sets of formulas are given to estimate the parameters of neural networks as neutrosophic triplets for weights and activations. The use of the NDNN in soil microbe processing is demonstrated to prove its usefulness in handling situations with uncertainty and indeterminacy.