Training Triplet Networks with GAN
This work addresses enhancing representation learning for classification and retrieval tasks, but it appears incremental as it combines existing techniques (triplet networks and GANs) without introducing a fundamentally new approach.
The authors tackled the problem of improving triplet network performance by training them as discriminators in GANs, resulting in significant classification improvements on Cifar10 and MNIST datasets using k-nn.
Triplet networks are widely used models that are characterized by good performance in classification and retrieval tasks. In this work we propose to train a triplet network by putting it as the discriminator in Generative Adversarial Nets (GANs). We make use of the good capability of representation learning of the discriminator to increase the predictive quality of the model. We evaluated our approach on Cifar10 and MNIST datasets and observed significant improvement on the classification performance using the simple k-nn method.