15 Relational Planning, Learning, and Probabilistic Reasoning

15.5 References and Further Reading

The situation calculus was proposed by McCarthy and Hayes [1969]. The form of the frame axioms presented here can be traced back to Kowalski [2014], Schubert [1990], and Reiter [1991]. Reiter [2001] presents a comprehensive overview of the situation calculus; see also Brachman and Levesque [2004]. The event calculus was proposed by Kowalski and Sergot [1986]. There have been many other suggestions about how to solve the frame problem, which is the problem of concisely specifying what does not change during an action. Shanahan [1997] provides an excellent introduction to the issues involved in representing change and to the frame problem in particular.

For overviews of inductive logic programming see Muggleton and De Raedt [1994], Muggleton [1995], and Quinlan and Cameron-Jones [1995].

The Netflix prize, and the winning algorithms are described at http://www.netflixprize.com/. The collaborative filtering algorithm is based on Koren et al. [2009]. The MovieLens data sets are described by Harper and Konstan [2015] and available from http://grouplens.org/datasets/movielens/.

Statistical relational AI is described by De Raedt et al. [2016]. Plate models are due to Buntine [1994], who used them to characterize learning. Independent choice logic was proposed by Poole [1993, 1997] and implemented in Problog [De Raedt et al., 2007]. De Raedt et al. [2008] and Getoor and Taskar [2007] provide collections of papers that provide overviews on probabilistic relational models and how they can be learned. Domingos and Lowd [2009] discuss how (undirected) relational models can provide a common target representation for AI.