Try the model without the similarities

[alexanderpanchenko]

Generate results similar to https://docs.google.com/spreadsheets/d/1esXh-eNz76_86PikQe9-FwkeEpMz7Of6rzeiSS6lrD8/edit#gid=1778525383, but ONLY using L2 regularization and regularization using nearest neighbours (the 'main' training data are not used at all in the form of the pair similarities).

[akutuzov]

...and with using the main training data and the nearest neighbors regularization (but without L2 or L1).

[m-dorgham]

@alexanderpanchenko what do you mean by "the 'main' training data are not used at all in the form of the pair similarities"?? Could you elaborate please.

[alexanderpanchenko]

I mean using only the 'regularizers' as the objective (the ones you added in your work).

[m-dorgham]

Here are the results: https://docs.google.com/spreadsheets/d/1esXh-eNz76_86PikQe9-FwkeEpMz7Of6rzeiSS6lrD8/edit#gid=2028670185

[alexanderpanchenko]

probably it will not change much, but just as a santy check please re-train also with beta=gamma=1.0

On Sep 19, 2018, at 3:39 PM, Mohammad Dorgham notifications@github.com wrote:

Here are the results: https://docs.google.com/spreadsheets/d/1esXh-eNz76_86PikQe9-FwkeEpMz7Of6rzeiSS6lrD8/edit#gid=2028670185 https://docs.google.com/spreadsheets/d/1esXh-eNz76_86PikQe9-FwkeEpMz7Of6rzeiSS6lrD8/edit#gid=2028670185 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/uhh-lt/path2vec/issues/14#issuecomment-422807486, or mute the thread https://github.com/notifications/unsubscribe-auth/ABY6vqB4KvFdB9o4bFXsEnnb0l4Y7PDDks5uckkugaJpZM4WlgK8.

[m-dorgham]

Ok. Will do it.