label_ranking_average_precision_score fails with sublists that are different sizes #8487
Comments
|
What do you intend to mean by that y_true?
…On 2 March 2017 at 08:30, José P. González-Brenes ***@***.***> wrote:
The evaluation function label_ranking_average_precision_score fails when
the sublists are of different sizes. Although I understand that this may be
out of implementation convenience, I think this should be documented or at
least a more adequate error should be raised.
import numpy as np
from sklearn.metrics import label_ranking_average_precision_score
y_true = np.array([[1, 0, 0], [0, 1]]) #3 and 2
y_score = np.array([[0.75, 0.5, 1], [1, 0.2, 0.1]]) # 3 and 2
label_ranking_average_precision_score(y_true, y_score)
This code fails with an exception:
ValueError: setting an array element with a sequence.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#8487>, or mute the
thread
<https://github.com/notifications/unsubscribe-auth/AAEz60AJnjuSi045zOkTC9K1jkl3uIAnks5rheN8gaJpZM4MQOkU>
.
|
|
Hello @jnothman, Not sure I understand your question. y_true is the "correct" answer for every search ranking opportunity. Every sublist is the number of results the user saw. Sometimes the user may see 3 search results, sometimes she may see 2 results, or whatever number. Currently, this metric only works when the user sees the same number of results in every session. If not, a cryptic error is raised. Thanks! |
803C
|
Scikit-learn doesn't really deal with learning-to-rank problems atm. I'm
inclined to close this issue as out of scope.
…On 2 March 2017 at 11:44, José P. González-Brenes ***@***.***> wrote:
Hello @jnothman <https://github.com/jnothman>,
Not sure I understand your question. y_true is the "correct" answer for
every search ranking opportunity. Every sublist is the number of results
the user saw. Sometimes the user may see 3 search results, sometimes she
may see 2 results, or whatever number.
Currently, this metric only works when the user sees the same number of
results in every session. If not, a cryptic error is raised.
Thanks!
Jose
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#8487 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAEz6wuOAL6iSMuzCTFysmhbO4p_iey8ks5rhhD_gaJpZM4MQOkU>
.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The evaluation function label_ranking_average_precision_score fails when the sublists are of different sizes. Although I understand that this may be out of implementation convenience, I think this should be documented or at least a more adequate error should be raised.
This code fails with an exception:
The text was updated successfully, but these errors were encountered: