Embedding learning constrains the searching space $\mathcal{H}$ by mapping each input $x_i \in \mathbb{R}^n$ to a lower-dimensional space $z_i \in \mathbb{R}^m$, where $m < n$. The mapping process is achieved by an embedding function $f(\cdot)$. We can achieve better accuracy by using different embedding functions $f_{\text{train}}$ and $f_{\text{test}}$ for the training and testing sets. The prediction on the testing set is made by finding the closest training sample to a given testing sample in the embedded space and predicting the same label for the testing sample as the training sample.