Original Model (Univariate)

One feature (house size):

fw,b(x) = wx + b

New Model (Multiple Features)

Many features:

Feature Variable Example
Size (sq ft) x1 2104
# Bedrooms x2 5
# Floors x3 1
Age (years) x4 45

More features = more information to predict price.

New Notation

Symbol Meaning Example
xj The jth feature x3= # of floors
n Total number of features n = 4
x(i) All features of the ith training example (a vector) x(2)= [1416, 3, 2, 40]
xj(i) Feature j of training example i x3(2)= 2 (2nd example, 3rd feature)

Note: Bold notation indicates a vector (a list of numbers), not a single number.

The Multiple Linear Regression Model

f,b(x) = w1x1 + w2x2 + w3x3 + w4x4 + b

General form with n features:

f,b(x) = w1x1 + w2x2 + ⋯ + wnxn + b

Concrete Example: Housing Prices

price = 0.1x1 + 4x2 + 10x3 − 2x4 + 80

Interpreting the parameters (price in $thousands):

Parameter Value Interpretation
b 80 Base price: $80,000
w1 0.1 +$100 per square foot
w2 4 +$4,000 per bedroom
w3 10 +$10,000 per floor
w4 -2 -$2,000 per year of age

Vector Notation

Define vectors to simplify the model: