Why does the Two-Stage-Method perform better than DART?
zytcs opened this issue · 2 comments
What's the intuition of the Two-Stage-Method? Can you provide more detail information about your method? Thanks in advance!
Hi. To be short, the iteration equation in DART tends to under-estimate the specificity of sources, leading low rejection rates for the answer provided by the source. The case is acceptable when the data cleaning is good and all sources provide reasonable answers (instead of garbles or irrational ones). However, if there is noise in sources, the perform of DART would be severely harmed. This result is proved in Xueling's paper (seeing the noise added experiment in her paper) and also my experiments. As for details about Two-Stage, I have attached my bachelor thesis under this repo. You can find details in the chapter regarding Two-Stage and experiments. Thank you.
Thank you very much:)