Linear regression models a continuous response as a linear function of one or more explanatory variables. It can describe how the response changes with its predictors and predict responses for future observations. section 3.4.1
For one explanatory variable, the response yᵢ is expressed using an intercept β₀, slope β₁, predictor xᵢ, and unexplained component εᵢ. section 3.4.1
The slope describes the expected response change for a one-unit predictor increase; the intercept gives the expected response when the predictor is zero. section 3.4.1
The fitted model uses estimated intercept and slope values to produce a predicted response for each observation. section 3.4.1
A residual is the difference between an observed response and its fitted prediction; residual size indicates their disagreement. section 3.4.1
Ordinary least squares chooses coefficients to minimize the sum of squared residuals. The aim is to capture the overall relationship, not to pass through every observation. section 3.4.1
With normally distributed errors, this geometric optimization and probabilistic estimation yield the same solution. The connection illustrates how maximum likelihood can unify estimation methods and links regression’s foundation to learning in neural networks. section 3.4.4 section 3.4.4



























