Reviews: Matching Networks for One Shot Learning

Comments: the idea contained in the paper is not a very big contribution to the field but still remains interesting, and the experiments on different datasets provide important information concerning the behavior of the model. But the paper has many different problems that make it difficult to understand. The first problem is in the structure of the paper since the problem formulation is in fact only given in Section 2.2 while I think it is important to well define the one-shot learning problem at the beginning of the paper. It would make the article easier to read and to follow. Sections 2.1 and 2.1.1 are well written but could include more precise definitions.