mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Abigail Gertner <gert...@mitre.org>
Subject Re: Computing userSimilarity in Taste AbstractSimilarity
Date Thu, 30 Sep 2010 21:53:43 GMT
 I think I must be looking at an older version of the file. I have
mahout-0.3 (the most recent one) downloaded from sourceforge. Maybe it
is updated since then in SVN?


On 9/30/2010 5:38 PM, Sean Owen wrote:
> I think it does work, but this code is definitely hard to grok. In my
> defense it is complex for a reason at least -- performance.
>
> When the end of one list of prefs is reached (the line "if (++xPrefIndex >=
> xLength)") it does check for an inferrer in the next line. If there is one,
> it sets "xIndex = Long.MAX_VALUE", and continues. This ensures the rest of
> the y values are compared against an inferred x value.
>
> Step through to convince yourself. Or, if you find I am actually completely
> wrong do let me know.
>
> On Thu, Sep 30, 2010 at 10:20 PM, Abigail Gertner <gertner@mitre.org> wrote:
>
>> Hello - I noticed something that I think might be a problem with the
>> userSimilarity computation in the AbstractSimilarity class. After
>> updating the running sums, the method checks the value of compare and
>> moves to the next preference value in the list that has the smaller item
>> index, or both if they are both on the same item. It breaks out of the
>> loop if it is at the end of the preference list(s) that it needs to get
>> the next value of. The problem is that it doesn't take into account if
>> there is a preference inferrer -- if it is at the end of one of the
>> lists, it will break out of the loop even if there is a preference
>> inferrer. This results in ignoring the rest of the items in the other
>> users preferences.
>>
>> Am I missing something or does this need to be corrected?
>>
>> Thanks for your help,
>> -Abigail
>>
>>

Mime
View raw message