I'm considering a leaf of a decision tree that consists of object-label pairs $(x_{1}, y_{1}), \dots, (x_{n}, y_{n})$.
The prediction $\hat{y}$ of this leaf is defined to minimize the loss on the training samples.
I have to find the optimal prediction in the leaf for a classification tree for $K$ classes, i.e. $y_{i} \in \{1, \dots, K\}$, and zero-one loss $\mathcal{L}(y, \hat{y}) = \left[y \neq \hat{y} \right]$.
This is all that is given in the task. Does anyone have an idea how to approach the task?