Editing Open Problems:86
Warning: You are not logged in. Your IP address will be publicly visible if you make any edits. If you log in or create an account, your edits will be attributed to your username, along with other benefits.
The edit can be undone.
Please check the comparison below to verify that this is what you want to do, and then save the changes below to finish undoing the edit.
Latest revision | Your text | ||
Line 1: | Line 1: | ||
{{Header | {{Header | ||
+ | |title=Equivalence testing lower bound via communication complexity | ||
|source=focs17 | |source=focs17 | ||
|who=Clément Canonne | |who=Clément Canonne | ||
}} | }} | ||
− | Blais, Canonne, and Gur {{cite|BlaisCG-17}} recently described a reduction technique to obtain distribution testing lower bounds from communication complexity (specifically, even from the simultaneous message-passing (SMP) setting in communication complexity). Using this technique, which is the analogue of the | + | Blais, Canonne, and Gur {{cite|BlaisCG-17}} recently described a reduction technique to obtain distribution testing lower bounds from communication complexity (specifically, even from the simultaneous message-passing (SMP) setting in communication complexity). Using this technique, which is the analogue of the "usual property testing" framework of Blais, Brody, and Matulef {{cite|BlaisBM-12}}, they prove and (re)-derive lower bounds for many of the standard distribution testing questions, including ''identity testing'' (and ''instance-specific'' identity testing — see [[Instance-specific Hellinger testing|this other open problem]]). |
Recall that the ''equivalence'' testing problem (also known as closeness testing) is the generalization of identity testing where, instead of having a known reference distribution $q$ and access to samples from an unknown $p$, now both distributions are unknown and only available via samples. | Recall that the ''equivalence'' testing problem (also known as closeness testing) is the generalization of identity testing where, instead of having a known reference distribution $q$ and access to samples from an unknown $p$, now both distributions are unknown and only available via samples. | ||
− | Can one use the {{cite|BlaisCG-17}} framework to re-establish an $\tilde{\Omega}(n^{2/3})$ sample complexity lower bound for equivalence testing? If so, can it yield some sort of ''instance-specific'' equivalence testing lower bound, for some meaningful notion of | + | Can one use the {{cite|BlaisCG-17}} framework to re-establish an $\tilde{\Omega}(n^{2/3})$ sample complexity lower bound for equivalence testing? If so, can it yield some sort of ''instance-specific'' equivalence testing lower bound, for some meaningful notion of "instance-specific"? |