Related to #6. Currently, all sequences of a given size are submitted to enumerate_frequent_seq(). This is slow and memory intensive. Memory requirements can be reduced in SPADE using a depth-first search constrained to a single equivalence class (a set of subsequences with a common n-prefix). This way, the entire lattice does not need to be held in memory at once -- only smaller sub-lattices need be available.
Modify use of enumerate_frequent_seq() so that it's called on individual equivalence classes.
Related to #6. Currently, all sequences of a given size are submitted to enumerate_frequent_seq(). This is slow and memory intensive. Memory requirements can be reduced in SPADE using a depth-first search constrained to a single equivalence class (a set of subsequences with a common n-prefix). This way, the entire lattice does not need to be held in memory at once -- only smaller sub-lattices need be available.
Modify use of enumerate_frequent_seq() so that it's called on individual equivalence classes.