-
-
Notifications
You must be signed in to change notification settings - Fork 53
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
MAE and RMSE for SLIM #14
Comments
SLIM based approaches are top-N recommendation algorithms. They cannot be
evaluated by MAE or RMSE
On Sun, Dec 17, 2017 at 02:48 Huyenntn ***@***.***> wrote:
I configured tem.ranking=off -topN 10 but results of SLIM was still Pre,
Rec,...
How can I get MAE and RMSE values?
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#14>, or mute the thread
<https://github.com/notifications/unsubscribe-auth/AHDB50lNP9jRYpBtzDHxoRQUXWXkVAiPks5tBDOigaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
Can you tell me what the config "-minlenu 2 -minleni 2" mean in usersplitting, itemsplitting and uisplitting algorithms. |
Let's take item splitting for example, ratings canbe split into two sets:
weekend and weekday. Next, we are going to use statistics to evaluate
whether there is significant different in ratings in these two contexts
The minlenu defines the minimal length of the set. If the size is too
small, the result by statistic is not reliable
In other words, if you increase the minimal length, the user or item will
be split when the two sets meet the length requirement and pass the
significance test
On Thu, Dec 21, 2017 at 15:23 Huyenntn ***@***.***> wrote:
Can you tell me what the config "-minlenu 2 -minleni 2" mean in
usersplitting, itemsplitting and uisplitting algorithms.
Thankyou very much!
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#14 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AHDB557RY4YRpYKM7WE-s30WuemViYrkks5tCiqhgaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
Thankyou and sorry to bother. I have another question. |
If u run userKNN, it will make predictions without considering context
For lc parameters, you need to read the paper or source codes
On Thu, Dec 21, 2017 at 20:06 Huyenntn ***@***.***> wrote:
Thankyou and sorry to bother. I have another question.
When I used DePaulMovie dataset with UserKNN. It's is a traditional
recommender system. But but the context still existed in the result.
And the second question is what does "-lc1 -lc2 and reg.lambda=0.0001 -c
0.001" in file config mean?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#14 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AHDB5zeRnShdjk4GK2zLT6NmtbfuJ1hJks5tCmz1gaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
For example with predictions in UserKNN-top-10-items fold [1].txt: |
And if it doesn't use the context, a vector (userid, itemid) may be duplicated. How did you solve it? |
We use average rating for user item pair
On Thu, Dec 21, 2017 at 21:17 Huyenntn ***@***.***> wrote:
And if it doesn't use the context, a vector (userid, itemid) may be
duplicated. How did you solve it?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#14 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AHDB58VtEB-8dRsUPk7qzvefcHBqkJcYks5tCn2KgaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
Can you give me the best config content for cslim_mcs algorithm with movie depaul data. I try it but evaluation value too bad. |
I don't have that at hand. For SLIM based algorithms, they are sensitive to
initializations and running parameters. You may need to carefully to tune
up the parameters
On Sat, Dec 23, 2017 at 11:17 Huyenntn ***@***.***> wrote:
Can you give me the best config content for cslim_mcs algorithm with movie
depaul data. I try it but evaluation value too bad.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#14 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AHDB5xpZYXAcWJ4nT33ObrqPViUqExJdks5tDJPogaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
Hello Yong Zheng. You still haven't answered my question. |
The outputs include contexts for each line. But as I said previously,
userknn will be run without considering contexts
On Sat, Dec 30, 2017 at 06:12 Huyenntn ***@***.***> wrote:
Hello Yong Zheng. You still haven't answered my question.
UserKNN make predictions without considering context. Why the context
still existed in the result?
Example of the result: https://i.imgur.com/kiLX9eg.png
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#14 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AHDB5yfYxNjtAEXRZKMj40hvFoqyqrlvks5tFii_gaJpZM4REdhr>
.
--
Sent from Gmail Mobile
|
I configured tem.ranking=off -topN 10 but results of SLIM was still Pre, Rec,...
How can I get MAE and RMSE values?
The text was updated successfully, but these errors were encountered: