Perceptron — shared thought trail

Historical context

Frank Rosenblatt introduced the perceptron in the late 1950s and developed the Mark I Perceptron for image-recognition experiments. The limits of simple perceptrons later tempered early enthusiasm, but the connectionist approach became foundational to neural networks.[1]

How it worked

A perceptron multiplies each input by a learned weight, adds the values and a bias, then applies a step function to produce a binary output. During training, an incorrect prediction triggers weight adjustments that make the desired response more likely.[1] This process finds a separating hyperplane when the classes are linearly separable.