Reviewer 1,Q1: Updates regarding background

[Suvodeep90]

Background. The paper builds upon a number of previous studies. However, several key questions have no answer: What is the bellwether? Why is this relevant? How does it work when applied to different contexts? Why is this useful in practice? None of these questions are addressed in the paper. This submission should be self-contained and allow a reader to understand its content without reading 10 other papers; in this sense, the submission fails completely. For example, the paper mentions that: "The core intuition of this new approach is that if many projects are similar, then we do not need to run comparisons between all pairs of projects": very good, but where this intuition come from?

1) What is* the bellwether?
2) Why is* this relevant?
3) How does it work when applied to different contexts?
4) Why is this useful in practice?
5) The core intuition of this new approach is that if many projects are similar, then we do not need to run comparisons between all pairs of projects: very good, but where this intuition come from? 

[Suvodeep90]

What is the bellwether? Bellwether is a source selection method for transfer learning. For any transfer learning process to transfer the learning to a target project, we need to select a suitable source project. These transfer learning processes are complex and computationally expensive in nature, thus it is impossible to run a transfer learner and see which one works best from a large pool of source projects. Bellwether solves this problem by selecting an exemplary source project from the source pool to initiate the transfer learning process.

[Suvodeep90]

Why is this relevant? Relevant because of transfer learning methods are expensive and for finding a source project to transfer we need to check all possible projects in the source pool. Also for normal transfer learning, we might different source projects for each target project thus introducing conclusion instability, using bellwether we can try to mitigate the conclusion instability.

[Suvodeep90]

How does it work when applied to different contexts? krishna et al(TSE paper) have shown that bellwether is applicable for multiple domains in software analytics. In this paper, we try to explore one of those domains (defect prediction) and come up with a better algorithm (faster)

[Suvodeep90]

Why is this useful in practice? 1) A large number of projects available in the open-source community 2) A project which looks similar may not be actually similar when doing prediction. 3) Mitigate conclusion instability