dm.cs.tu-dortmund.de/mlbits/frequent-pattern-acceleration/
Scaling Frequent Itemset Mining – Lecture Notes
2\) pairs
for triples, we usually do not have to count \(M^3\) triples because of Apriori pruning
(in the Netflix example, we even had \(|C_3|\lt{}|C_2|\) !)
Counting All Pairs
Assume we have 100.000 products [...] matrix wins
But beware of memory management (and programming language) overheads!
theoretical
Py2
Py3
numpy
R
Java
(opt.)
C (STL)
(i,j,c)
12
152
168
120
72
24
24
16
set (i,j,c)
\(\approx 20\)
\(\approx …
