This Paper is to solve the transfer Learning Problem.
Keywords: Adversarial loss, Metric, Sparse Label,
tease： From the abstract, the paper is totally result oriented and application focused. (Our method shows compelling results on novel classes within a new domain even when only a few labeled examples per class are available, outperforming the prevalent fine-tuning approach. In addition, we demonstrate the effectiveness of our framework on the transfer learning task from image object recognition to video action recognition.)
I just finished the Multi-layer domain transfer, the key to is make the transform of each layer. This methods seems self-evident, I need more time and equations to double check this.
Let's assume this is true. Then we can make the transfer from the trained feature space to target feature space.
All we left is the final layer output. The metric-based cross entropy loss require math and calculation. [Todo]
