A fully connected layer (FC layer), also known as a dense layer, in a neural network is a layer where each neuron is connected to every neuron in the previous layer. This means it takes a vector input (flattening any structure from previous layers) and produces a vector output. Each output is a weighted sum of all inputs plus a bias, followed by a non-linear activation (if not the last layer). FC layers are typically found at the end of CNN architectures to integrate all the extracted features and make a final prediction (e.g., class scores). However, purely fully connected networks (multi-layer perceptrons) can also be the whole model for tabular data. FC layers have a lot of parameters (weights), which is manageable for moderate-sized inputs but not for large images unless after pooling/flattening. They are good at capturing global interactions between features but don’t exploit spatial structure (hence CNNs use conv layers for most of the work, then FC for final decision).