what are cats?

[drakeeee]

What are the cats argument required in the LogisticRegression model for example? I can't find any description of its identity in your documentation. Is it the number of categories to infer per sample?

[danielemaasit]

@drakeeee Yes, it is the number of categories.

@arteagac can you make the documentation more clear on the description of this argument.

[Emaasit]

pymc-learn/pmlearn/linear_model/logistic.py

Line 24 in 4f1ee68

self.num_cats = None

[drakeeee]

OK. The documentation or examples don't reveal much to what cats should be.
For example see:
http://docs.pymc-learn.org/en/latest/notebooks/LogisticRegression.html

where the data is defined:

x_a = np.random.randn(num_samples, num_pred)
y_a = np.random.binomial(1, numpy_invlogit(alphas[0] + np.sum(betas[0] * x_a, 1)))
x_b = np.random.randn(num_samples, num_pred)
y_b = np.random.binomial(1, numpy_invlogit(alphas[1] + np.sum(betas[1] * x_b, 1)))

X = np.concatenate([x_a, x_b])
y = np.concatenate([y_a, y_b])
cats = np.concatenate([
    np.zeros(num_samples, dtype=np.int),
    np.ones(num_samples, dtype=np.int)
])

Maybe I am being dense but I don't understand from this snippet whats cats is or should be in a logreg model.

[drakeeee]

From that example it appears that you need to know cats before testing. Let me know if I am missing some kind of understanding.