You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by yamo93 <ya...@gmail.com> on 2012/10/30 15:55:00 UTC
Performance : user-based vs item-based recommenders
Hi,
My input file contains more items than users, so according to the book
MAI, i expect that user-based recommendation is faster than item-based
recommendation but it's not what i see : item-based is 5 times faster.
Here's some characteristics of my input file
Nb users:258
Nb items:56756
Nb lines:344816
Have you some explanations ?
Thks,
Yann.
Re: Performance : user-based vs item-based recommenders
Posted by yamo93 <ya...@gmail.com>.
Thanks for reply.
I only use the constructor GenericItemBasedRecommender with 2 params and
i just discovered the possibilities offered by the 4 params ctor... I
will investigate.
On 10/30/2012 04:34 PM, Sean Owen wrote:
> Well what code are you running?
> Look at the source and look specifically at CandidateItemStrategy's usages.
>
> On Tue, Oct 30, 2012 at 3:19 PM, yamo93 <ya...@gmail.com> wrote:
>
>> On which criteria items are pruned ?
>> What kind of further information do you need ?
>>
>>
Re: Performance : user-based vs item-based recommenders
Posted by Sean Owen <sr...@gmail.com>.
Well what code are you running?
Look at the source and look specifically at CandidateItemStrategy's usages.
On Tue, Oct 30, 2012 at 3:19 PM, yamo93 <ya...@gmail.com> wrote:
> On which criteria items are pruned ?
> What kind of further information do you need ?
>
>
Re: Performance : user-based vs item-based recommenders
Posted by yamo93 <ya...@gmail.com>.
On which criteria items are pruned ?
What kind of further information do you need ?
On 10/30/2012 03:59 PM, Sean Owen wrote:
> Probably because the item-based one prunes items from consideraiton by
> default. You'd have to provide a lot more information to say more.
>
> On Tue, Oct 30, 2012 at 2:55 PM, yamo93 <ya...@gmail.com> wrote:
>
>> Hi,
>>
>> My input file contains more items than users, so according to the book
>> MAI, i expect that user-based recommendation is faster than item-based
>> recommendation but it's not what i see : item-based is 5 times faster.
>>
>> Here's some characteristics of my input file
>> Nb users:258
>> Nb items:56756
>> Nb lines:344816
>>
>> Have you some explanations ?
>>
>> Thks,
>> Yann.
>>
Re: Performance : user-based vs item-based recommenders
Posted by Sean Owen <sr...@gmail.com>.
Probably because the item-based one prunes items from consideraiton by
default. You'd have to provide a lot more information to say more.
On Tue, Oct 30, 2012 at 2:55 PM, yamo93 <ya...@gmail.com> wrote:
> Hi,
>
> My input file contains more items than users, so according to the book
> MAI, i expect that user-based recommendation is faster than item-based
> recommendation but it's not what i see : item-based is 5 times faster.
>
> Here's some characteristics of my input file
> Nb users:258
> Nb items:56756
> Nb lines:344816
>
> Have you some explanations ?
>
> Thks,
> Yann.
>