SciPost Submission Page
Navigator Function for the Conformal Bootstrap
by Marten Reehorst, Slava Rychkov, David Simmons-Duffin, Benoit Sirois, Ning Su, Balt van Rees
This Submission thread is now published as
|Authors (as Contributors):||Marten Reehorst · Slava Rychkov · Balt van Rees|
|Date submitted:||2021-07-12 12:15|
|Submitted by:||van Rees, Balt|
|Submitted to:||SciPost Physics|
Current numerical conformal bootstrap techniques carve out islands in theory space by repeatedly checking whether points are allowed or excluded. We propose a new method for searching theory space that replaces the binary information “allowed”/“excluded” with a continuous “navigator” function that is negative in the allowed region and positive in the excluded region. Such a navigator function allows one to efficiently explore high-dimensional parameter spaces and smoothly sail towards any islands they may contain. The specific functions we introduce have several attractive features: they are well-defined in large regions of parameter space, can be computed with standard methods, and evaluation of their gradient is immediate due to an SDP gradient formula that we provide. The latter property allows for the use of efficient quasi-Newton optimization methods, which we illustrate by navigating towards the 3d Ising island.
Published as SciPost Phys. 11, 072 (2021)
Author comments upon resubmission
With this resubmission we have addressed most of the concerns of the referees. A full list is given below. One significant change is the addition of a new short subsection 3.1 where we provide evidence that the navigator function is not everywhere C^2.
In addition we would like to provide the following replies to some of the remarks in the reports.
Report 1 asked for more details of the Sigma navigator functions. However in this work we did not fully explore Sigma navigators. We added footnote 34 on page 42 to stress that at present both navigators are on equal footing. In our computations with a Sigma navigator some really simple minded choices worked well, like summing about 100 conformal blocks with small scaling dimensions spacing and equal coefficients. There was nothing particularly smart about those choices, so we preferred not to report them.
Report 1 also asked whether one would find a flat plane in the case of a conformal manifold. We do not expect this to be true, and would also like to point out that this does not logically follow from our claim that the minimum of the navigator corresponds to a CFT.
In report 1 a question was also raised about using the navigator for OPE space islands. In our view it makes no conceptual difference whether the parameter space in which one navigates consists of scaling dimensions, OPE coefficients, or both (as we do in the paper). Therefore we did not supply any further comments. Of course the authors would be happy to offer practical guidance in specific cases.
We expect the request in report 2 to "add a comment or plot" about the navigator inside the Ising island to be addressed by the new subsection 3.1.
List of changes
In response to report 1:
- Added footnote 34 on page 42;
- Changed "a good predictor for the location of the true CFT" -> "close to the true CFT" on page 25.
In response to report 2:
- In abstract: replace "everywhere well-defined" -> "well-defined in large regions of parameter space";
- Added footnote 33 on page 41 to compare the number of SDPB iteration between the navigator method and Delaunay triangulation method;
- Added and the value of the navigator at the minima in equations (5.8) on p31 and (5.9) on p35.
In reponse to report 3:
- We increased the size of black and red dots of figure 13 (previously figure 12) to make them more visible.
- Added "We used the same conformal block normalization as ." in appendix D on page 55;
- Corrected caption of figure 13 (previously figure 12): changed "red/black" -> "black/red", changed "allowed/disallowed" to "allowed/excluded";
- On page 23, in the acknowledgments on page 44, around and in footnote 41 on page 51: added information that the announced program is now available under the name "approx_objective" as a part of the SDPB package. Also changed "we than" -> "we thank" in footnote 41;
- Added subsection 3.1 and figure 5 to provide evidence for some non-C^2-differentiability of the navigator. Relatedly:
-- on p14: "The navigator is observed to be smooth" -> "On this scale the navigator is observed to be smooth (see however below)";
-- on p15: "smooth" -> "smooth at this scale";
-- on p33: added footnote 30 on plateaus;
-- on p42: "smooth" -> "(C^1) smooth".
Submission & Refereeing History
You are currently on this page
Reports on this Submission
Anonymous Report 3 on 2021-9-2 (Invited Report)
The revisions made by the authors are appreciated. This paper should be published in its current form.
Report 2 by Kay Joerg Wiese on 2021-8-3 (Invited Report)
The article should be published as is.
Anonymous Report 1 on 2021-7-15 (Invited Report)
The paper is now ready to be published.