Gradient Descent

To understand, consider simpler linear unit, where

$$o = w_{0} + w_{1}x_1 + \cdots + w_n x_n$$

Let's learn w_i's that minimize the squared error

$$E[\vec{w}] \equiv \frac{1}{2}\sum_{d \in D}(t_{d} - o_{d})^{2}$$

Where D is set of training examples