Disclaimer: Dieser Thread wurde aus dem alten Forum importiert. Daher werden eventuell nicht alle Formatierungen richtig angezeigt. Der ursprüngliche Thread beginnt im zweiten Post dieses Threads.

Hi SAKI Team,

please i have two questions:

1-) for this formular (cos(x,y) = <x,y> / |x| |y|) here: do we use L2 or L1 norm?

2-) (views@n) here to compute viewsAtN(sims, 50) is ‘50’ an userId or it is the number of recommendations ?


this is the cosine similarityÄhnlichkeit

Thanks!!! do you have any idea about my second question? Did you used the number “50” as an userId or as a number of recommendations ?

I think its the number of recommendations, but i am not at that point yet

I see that my calulation of the cosine simularity is very slowly. Do we really have to compare every movie with each other, what will take about 1 min per movie with my computation (resulting in about 7 days computing time).

I don’t know how you do your calculations and why it takes a lot of time for you, but I have applied function sklearn.metrics.pairwise.cosine_similarity on the transpose of the ratings matrix and it doesn’t take me so long to execute.

Sounds interesting, i modified my calculations so it is about factor 60 faster now (but it still takes a lot of time).
I will try it with your tipp, thank u