#mathematic #causality # [[Epistemic status]] #shower-thought #to-digest # Changelog ```dataview TABLE WITHOUT ID file.mtime AS "Last Modified" FROM [[#]] SORT file.mtime DESC LIMIT 3 ``` # Related [[Graph neural networks]] [[thegradient.pub - Transformers Are Graph Neural Networks|Transformers Are Graph Neural Networks]] [[Geometric deep learning]] [[A Star path finding is written in nature]] # TODO > [!TODO] TODO > applications > a* > gnn > business applications # Graph >Both directed and undirected **graphs** have been used by researchers to facilitate such decomposition. Undirected **graphs**, sometimes called [[Markov Network|Markov network]]s (Pearl 1988b), are used primarily to represent symmetrical spatial relationships (Isham 1981; Cox and Wermuth 1996; Lauritzen 1996). Directed **graphs**, especially **DAGs**, have been used to represent **causal or temporal relationships** (Lauritzen 1982; Wermuth and Lauritzen 1983; Kiiveri et al. 1984) and came to be known as **Bayesian networks**, a term coined in Pearl (1985) to emphasize three aspects: (1) the subjective nature of the input informa tion; (2) the reliance on [[Bayes theorem]]'s conditioning as the basis for updating [[Information|information]]; and (3) the distinction between causal and evidential modes of reasoning, a distinction that underscores Thomas [[Bayes theorem]]'s paper of 1763. > Hybrid **graphs** (involving both directed and undirected edges) have also been proposed for statistical modeling (Wermuth and Lau ritzen 1990), but in this book our main interest will focus on directed acyclic **graphs**, with occasional use of directed cyclic **graphs** to represent feedback cycles. Undirected -> often used in [[Reinforcement Learning]]? DAG -> [[Mathematic/Causality/Causality]] Hydrid -> uncertainty, [[Probability]]