Getting topN arm features in onlinepolicy method

[Shriaachar]

topN arms for all online policy's not available. Results obtained cannot be mapped with requirements

[david-cortes]

Could you be more specific?

[Shriaachar]

After applying online policy's we are not able to get the topN products . its throws an error.

[david-cortes]

I'm unable to find any issue. The following code for example works for me:

import numpy as np
from sklearn.linear_model import LogisticRegression
from contextualbandits.online import EpsilonGreedy

rng = np.random.default_rng(seed=123)
nrows = 100
ncols = 10
narms = 5
X = rng.standard_normal(size=(nrows,ncols))
a = rng.integers(narms, size=nrows)
r = (rng.random(size=nrows) <= .3).astype(int)

policy = EpsilonGreedy(
    LogisticRegression(),
    narms
)
policy.fit(X,a,r)
policy.topN(X, n=3)

Are you able to provide a reproducible example with random data resulting in some error being thrown?

[Shriaachar]

AttributeError Traceback (most recent call last) Input In [1], in <cell line: 18>() 13 policy = EpsilonGreedy( 14 LogisticRegression(), 15 narms 16 ) 17 policy.fit(X,a,r) ---> 18 policy.topN(X, n=3)

AttributeError: 'EpsilonGreedy' object has no attribute 'topN'

I am getting this error

[david-cortes]

I am guessing that you are using an older version of this package that didn't support such functionality. Try updating:

pip install -U contextualbandits

[Shriaachar]

Getting requirements to build wheel: started Getting requirements to build wheel: finished with status 'done' Preparing wheel metadata: started Preparing wheel metadata: finished with status 'done' Requirement already satisfied: joblib>=0.13 in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (1.1.0) Requirement already satisfied: scipy in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (1.7.3) Requirement already satisfied: numpy>=1.17 in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (1.21.5) Requirement already satisfied: scikit-learn>=0.22 in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (1.1.3) Requirement already satisfied: pandas>=0.25.0 in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (1.4.2) Requirement already satisfied: cython in c:\users\shridhar.s\anaconda3\lib\site-packages (from contextualbandits) (0.29.28) Requirement already satisfied: pytz>=2020.1 in c:\users\shridhar.s\anaconda3\lib\site-packages (from pandas>=0.25.0->contextualbandits) (2021.3) Requirement already satisfied: python-dateutil>=2.8.1 in c:\users\shridhar.s\anaconda3\lib\site-packages (from pandas>=0.25.0->contextualbandits) (2.8.2) Requirement already satisfied: six>=1.5 in c:\users\shridhar.s\anaconda3\lib\site-packages (from python-dateutil>=2.8.1->pandas>=0.25.0->contextualbandits) (1.16.0) Requirement already satisfied: threadpoolctl>=2.0.0 in c:\users\shridhar.s\anaconda3\lib\site-packages (from scikit-learn>=0.22->contextualbandits) (2.2.0) Building wheels for collected packages: contextualbandits Building wheel for contextualbandits (PEP 517): started Building wheel for contextualbandits (PEP 517): finished with status 'error' Failed to build contextualbandits

See it's failed to build contextualbandits

[david-cortes]

So it's a version problem then. The installation error trace unfortunately doesn't show what the exact error is, so I cannot help with that.

Perhaps adding argument --verbose or something like that could help in showing an error trace, but I'm no expert on pip. There's also the possibility of using python setup.py install if you download the repository, which might provide some hint of what is failing.

[Shriaachar]

Hi David Every time I ran the topN I am getting different recommendations.

How can we resolve this ?

[david-cortes]

As it should be. You can check the docs and references for more info.