In neural network models, neurons process input features to produce an output based on a weighted sum and an activation function. Some neurons are designed to accept binary inputs, where each input feature \( x_i \in \{0, 1\}\). Consider the following neuron types:
- McCulloch-Pitts Neuron: An early computational model (1943) that uses binary inputs and outputs, designed to mimic logical operations with a threshold activation.
- Perceptron: A single-layer model (1958) introduced by Rosenblatt, capable of learning linear decision boundaries, typically accepting continuous or binary inputs but often using binary inputs in early applications.
- Modern Neuron: A neuron in contemporary neural networks, typically part of deep networks, processing continuous real-valued inputs with activation functions like ReLU or sigmoid.
Question:
Which neuron type is most closely associated with using strictly binary inputs?