Birds Fine-grained Feature Extraction Based on Transfer Learning

Abstract

The KLt-SNE algorithm is based on Kullback-Leibler divergence and t-SNE. Give an unknown distribution $p(x)$, at first, and then establish a $q\, (x\vert \theta)$, with the same dimension as the unknown distribution, estimate the parameter $\theta$ that needs to be configured by taking $N$ samples from $p(x)$. The results show that this algorithm can effectively achieve dimensionality reduction of data.