CStree data structure

[felixleopoldo]

As the CStrees seems to take O(p2^p) space for binary variables, which I think makes them intractable for larger p, maybe we can either use

1. The DAG sequence as internal data structure, which I think takes O((#minl stages) * 2^(max # parents)) space for the parameters and O((#minl stages) * p^2) for the graph.
2. Only store the colored stages and not the whole tree, as we already know what the tree looks like. E.g. for the CStree in Fig. 1 in Duarte & Solus 2022: [null, null, [{11, 12}], [{111,121}, {112, 122}, {211, 221}] ]. Which will also takes O(p2^p) space, but in practice maybe less depending on the number of stages.

[felixleopoldo]

As CStrees may take up too much space to store directly, which operations on CStrees do we want to implement?

[felixleopoldo]

1. There can be like O(2^p) different minimal contexts, so maybe it's impossible since there would be about one DAG per minimal context. Maybe we would need some limit on the number of nodes in the minimal contexts.