An FNN is the simplest neural network where connections do not form cycles. Information moves in one direction: from input, through hidden layers, to output.
The classic 'Multilayer Perceptron' architecture.
Foundational architecture before RNNs and Transformers.