The paper presents a very interesting and efficient technique for transfer learning which is validated on image classification data. The three reviewers agreed on the quality and siginficance of the contributions. I recommend acceptance as a poster presentation.