# Description The perceptron is a structure proposed by Rosenblatt (1962). Represent the most simple neural network. Although it does not hold significant historical relevance today, the [[Input or combination function]] and [[Activation function]] functions are as follows: - **Input Function:** The input function of the neuron is the weighted sum of the inputs $x_j$ and the weights $w_j^i$ $ f(x) = \sum_{j=1}^{n} x_j w_j^i $ - **Activation Function:** The activation function is represented using the **step function**. The corresponding output is: $ y_i = H(f^i(x)) = H(x_1 w_1^i + \dots + x_n w_n^i) $ $ y_i = \begin{cases} 0 & \text{if } x_1 w_1^i + \dots + x_n w_n^i - \alpha \leq 0 \\ 1 & \text{if } x_1 w_1^i + \dots + x_n w_n^i - \alpha > 0 \end{cases} $ where $\alpha$ is the threshold value. These neurons are called **Threshold Logic Units (TLUs)**. A perceptron consists of one or more TLUs organized in a single layer. --- ## References - [[Deep learning - Anna Bosch Rué Jordi Casas Roma Toni Lozano Bagén]]