Figure 2
From: Quantum adversarial metric learning model based on triplet loss function
The schematic of the metric learning model based on triplet loss function. A triplet set includes an anchor sample, a positive sample, and a negative sample. The input consists of a batch of triplet sets, and only one triplet set serves as input in each iteration. Three CNN with the same structure and parameters are used to map the triplet set into the embedded representation space. CNN, consisting of multiple convolutions, pooling, and fully connected layers, is responsible for extracting the features of samples. The triplet loss function is further constructed based on the extracted features