Outline
- building block of regression
- Q1: least squares
- aside: newton -raphson method
- Q2: min-norm least squares
- Q3: intro to softmax
Supervised Learning
| Classification | Regression |
|---|---|
| Spam/Ham | Housing prices |
| Dog/Cat | Stock Returns |
| One/ Two/ … Nine | Biophysical Measurements |
| Red Wine / White Wine |
Regression Function: Priors + Structure of Data
- Linear (y = mx)
- Polynomial
- Logistic
- Neural Net
Loss Functions: Statistical Model (Dis 5 + Lecture )
- Squared Error
- absolute error
- cross entropy
Cost Functions: Statistical Model (Dis 5 + Lecture )
- Mean / Sum / Max
- Regularization
Squard Error + Sum / Mean ←> () Cross Entropy + Sum / Mean (Multinormal and LR test)