I know from here that a log-linear model can be used to estimate a conditional probability of class $c$ given the feature representation $d$ of datapoint $x$.

$p(c|d;\theta) = \frac{exp(\theta.d_c)}{\sum_{d_{c^\prime}}exp(\theta.d_{c^\prime}) }$

How can I generalize the above formulation of log-linear models to estimate $p(c|d_1,d_2;\theta)$.


1 Answer 1


This is already multivariate. For $p$ features:

\begin{equation} \theta^\top.d_c = \sum_{i=1}^p\theta_i * {d_c}_i \end{equation}

  • $\begingroup$ So how to solve $p(c|d1,d2)$? $\endgroup$
    – MAZDAK
    Jun 23, 2016 at 5:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.