Similarities Between Adjusted R-Squared and AIC
Problem statement
When performing stepwise regression (forward or backward selection), you often use metrics like adjusted $R^2$ and AIC to compare candidate models.
What are the key similarities between adjusted $R^2$ and AIC as model selection criteria? In what ways do they differ?
Show hint
What raw goodness-of-fit quantity does each criterion build on, and how is it modified?
Relevant roles: Data Scientist, Applied Scientist and ML Engineer.