[MS0540] Mathematics for AI

Gradient Descent

$$Error = Y_{true} - Y_{predict}$$

Loss Function

$$COST = \sum_{i=1}^n (Y\_true_i - Y\_predict_i)^2$$

We want to minimize the cost of the function. we can re-model and substitute the equation above into a simpler term below.

Essentially,

$$f(x) = X^2, \\ y=X^2$$

What it means to minimize the function above is - to find the X value that can produce the lowest y-value. (the minimum point of the graph y=X^2)