# ML from Scratch — Episode 2: Linear Regression, The Implementation

## The Math

Linear regression learns to predict a continuous value from input data, by finding the best-fit line through the training examples.

### Cost Function

To measure how wrong our model is, we use MSE (Mean Squared Error):

$$J(w, b) = \frac{1}{2m} \sum_{i=1}^{m} (\hat{y}^{(i)} - y^{(i)})^2$$

Where m is the number of training examples, ŷ is the predicted value, and y is the actual label.We square the differences to penalize large errors more heavily-an error of 10 becomes 100, while an error of 2 becomes 4.

### Gradients

To minimize the cost, we compute how it changes with respect to each parameter:

$$\frac{\partial J}{\partial w} = \frac{1}{m} \sum_{i=1}^{m} (\hat{y}^{(i)} - y^{(i)}) \cdot x^{(i)}$$

$$\frac{\partial J}{\partial b} = \frac{1}{m} \sum_{i=1}^{m} (\hat{y}^{(i)} - y^{(i)})$$

### Gradient Descent

We update the parameters on every iteration:

$$w = w - \alpha \cdot \frac{\partial J}{\partial w}$$

$$b = b - \alpha \cdot \frac{\partial J}{\partial b}$$

Here α is the learning rate —> it controls how big each step is toward the minimum. Too large and you overshoot. Too small and training takes forever. Too large and you overshoot.

The goal: reduce **J** as much as possible with each iteration.