liveBook · Manning

AIC

As explained in Timeless Algorithms, The Foundational Ideas

The Akaike information criterion is a model selection criterion that estimates relative information loss while balancing fit against model complexity.

When competing models are not nested, likelihood-ratio tests may be inappropriate. AIC offers an alternative: it starts with a model’s maximized likelihood and penalizes complexity, favoring models whose improved fit justifies additional parameters. Lower values indicate a more favorable balance of fit and parsimony. section 3.7.2

AIC emphasizes predictive performance and is often preferred for forecasting future observations. Compared with AIC, BIC penalizes additional parameters more strongly, especially as sample size grows, and tends to favor simpler models. Both criteria use maximum likelihood for model comparison while guarding against overfitting. section 3.7.2

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Explains the maximized-likelihood and complexity-penalty structure and interprets lower scores, without supplying a formula.
What it assumes
Assumes fitted models and their maximized likelihoods, without indicating which prerequisites are taught from zero.
What you can do after
Evaluate competing models by balancing maximized likelihood against a complexity penalty and preferring lower AIC values.
How it connects
Connects maximum likelihood, model complexity, predictive performance, BIC, likelihood-ratio tests, and overfitting.

Also explained in

Compare how 4 books explain AIC →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.