diff --git a/exercises-pdf/ic_classification_1.pdf b/exercises-pdf/ic_classification_1.pdf
index 4d266eaf5..3d7500ae2 100644
Binary files a/exercises-pdf/ic_classification_1.pdf and b/exercises-pdf/ic_classification_1.pdf differ
diff --git a/exercises/supervised-classification/ex_rnw/ex_research-group-logreg.Rnw b/exercises/supervised-classification/ex_rnw/ex_research-group-logreg.Rnw
index c0d8a3fe3..d5223858c 100644
--- a/exercises/supervised-classification/ex_rnw/ex_research-group-logreg.Rnw
+++ b/exercises/supervised-classification/ex_rnw/ex_research-group-logreg.Rnw
@@ -17,7 +17,7 @@ Researcher Lisa knows that logistic regression follows a discriminant approach,
 \end{align}
 Additionally, she recalls the Bernoulli loss function of the logistic regression model in statistics:
 \begin{align}
-\Lpixy = -\;y\ln(\pix)-(1-y)\ln(1-\pix)
+\Lpixy = \lcrossent %-\;y\ln(\pix)-(1-y)\ln(1-\pix)
 \end{align}
 Lastly, she recollects how logistic regression models the posterior probabilities $\pixt$ of the labels $\text{--}$ the estimated linear scores are "squashed" through the logistic function $s$:
 \begin{align}