search

yingtaoluo / Spatial-Temporal-Attention-Network-for-POI-Recommendation

Codes for a WWW'21 Paper. POI recommender system for location/trajectory prediction.
https://doi.org/10.1145/3442381.3449998
164 stars 39 forks source link

A question about train_m2t. #10

Closed coolcoolcat closed 3 years ago

coolcoolcat commented 3 years ago

I have benefited a lot from reading your code and articles, but there is one problem I can't understand.

In line 187 and 188 of file train.py, read as follows: 'train_m2t = person_m2t[:, mask_len - 1] train_label = person_label[:, mask_len - 1]'. From this we can get the location of the tag ，and the time interval between the user accessing that location and the previous location.That is, the model already knows the time interval between the next visit and the current one，which is not reasonable for prediction.

I am not sure if there is something wrong with my understanding. Please kindly give me your advice. Thank you very much.

yingtaoluo commented 3 years ago

The model will be fed "time interval between current visit (last data in sequence) and next visit (label)", but no "distance between current location (last data in sequence) and next location (label)". The time interval can be useful information, but distance will cause a causality problem.

[image: image.gif]

On Wed, Aug 18, 2021 at 4:12 PM coolcoolcat @.***> wrote:

I have benefited a lot from reading your code and articles, but there is one problem I can't understand.

In line 187 and 188 of file train.py, read as follows: 'train_m2t = person_m2t[:, mask_len - 1] train_label = person_label[:, mask_len - 1]'. From this we can get the location of the tag ，and the time interval between the user accessing that location and the previous location.That is, the model already knows the time interval between the next visit and the current one，which is not reasonable for prediction.

I am not sure if there is something wrong with my understanding. Please kindly give me your advice. Thank you very much.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/yingtaoluo/Spatial-Temporal-Attention-Network-for-POI-Recommendation/issues/10, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH5PX43O5XTDPTKFTMCXQM3T5NTNHANCNFSM5CLPUX7Q . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email .

yingtaoluo commented 3 years ago

If we know "distance", we know the label location; if we know "interval", we just know when the user plans to move to the next location.

On Wed, Aug 18, 2021 at 4:12 PM coolcoolcat @.***> wrote:

I have benefited a lot from reading your code and articles, but there is one problem I can't understand.

In line 187 and 188 of file train.py, read as follows: 'train_m2t = person_m2t[:, mask_len - 1] train_label = person_label[:, mask_len - 1]'. From this we can get the location of the tag ，and the time interval between the user accessing that location and the previous location.That is, the model already knows the time interval between the next visit and the current one，which is not reasonable for prediction.

I am not sure if there is something wrong with my understanding. Please kindly give me your advice. Thank you very much.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/yingtaoluo/Spatial-Temporal-Attention-Network-for-POI-Recommendation/issues/10, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH5PX43O5XTDPTKFTMCXQM3T5NTNHANCNFSM5CLPUX7Q . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email .

coolcoolcat commented 3 years ago

If we know "distance", we know the label location; if we know "interval", we just know when the user plans to move to the next location. … On Wed, Aug 18, 2021 at 4:12 PM coolcoolcat @.***> wrote: I have benefited a lot from reading your code and articles, but there is one problem I can't understand. In line 187 and 188 of file train.py, read as follows: 'train_m2t = person_m2t[:, mask_len - 1] train_label = person_label[:, mask_len - 1]'. From this we can get the location of the tag ，and the time interval between the user accessing that location and the previous location.That is, the model already knows the time interval between the next visit and the current one，which is not reasonable for prediction. I am not sure if there is something wrong with my understanding. Please kindly give me your advice. Thank you very much. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#10>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH5PX43O5XTDPTKFTMCXQM3T5NTNHANCNFSM5CLPUX7Q . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email .

Thank you very much for your reply！