Typo in VBLogisticRegression model

[btlorch]

Hi there,

thank you for sharing your implementations! In the update of the variational distribution q(alpha) for the variational Bayesian logistic regression model, I believe there is a small mistake.

According to Eq. 10.179 in Bishop's book, we use the expectation of w^T w, and that is mu_N^T mu_N + trace(Sigma_N).
Therefore

(np.sum(w ** 2) + np.sum(Ri ** 2))

should be changed to

(np.sum(w ** 2) + np.trace(Ri ** 2))

and analogously in the line above.

Hope this helps!