Notes for <Show, Attend and Tell: Neural Image Caption Generation with Visual Attention>