search

baruchbarzel / NatureBiotech-31-720

2 stars 0 forks source link

Avoid singularity check? #1

Open milani opened 7 years ago

milani commented 7 years ago

In this line. R=N. Is it intentional to avoid re-normalizing values? Why?

baruchbarzel commented 7 years ago

Its to avoid the possibility that G is singular (we invert it later on). In principle a perfect G (from perturbation data) should not be singular. In practice, the user's noisy, partial G, which is obtained from correlation or mutual information data could be singular or close to singularity (in which case R < N).

The rationale is that in statistical correlations the specific magnitude of a term is not where the information lies, but rather in the relationship between terms, e.g. Gij > Gmn. This allows us some freedom to renormalize the off-diagonal terms of G without harming (too much) the info encoded within.

http://u.cs.biu.ac.il/~barzelb/ Baruch Barzel Complex Networks Dynamics @ Bar-Ilan University Ramat Gan Israel Homepage http://u.cs.biu.ac.il/~barzelb/index.html

On Tue, Oct 31, 2017 at 7:42 AM, Morteza Milani notifications@github.com wrote:

In this line https://github.com/baruchbarzel/NatureBiotech-31-720/blob/e84d4fcc18dc14fb1b5cee50dcfecb42f4b9cfc1/SupplementarySoftware1/SILENCING.m#L56. R=N. Is it intentional to avoid re-normalizing values? Why?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/baruchbarzel/NatureBiotech-31-720/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/ALDaHCGxbWrNXsLa-RiAJimfkzEGrBq-ks5sxrM7gaJpZM4QMPK8 .

milani commented 7 years ago

Problem is that the next if statement will be useless, since R = N the statement R < N will always be false.

baruchbarzel commented 7 years ago

OMG you are correct! Just checked my original code (on my PC) and this line R = N does NOT appear there. I dont know how it slipped into the public version, but should be removed. Hope I am not missing something - this code is from 2013... but pretty sure its a bug.

thanks for noticing.

http://u.cs.biu.ac.il/~barzelb/ Baruch Barzel Complex Networks Dynamics @ Bar-Ilan University Ramat Gan Israel Homepage http://u.cs.biu.ac.il/~barzelb/index.html

On Tue, Oct 31, 2017 at 9:22 AM, Morteza Milani notifications@github.com wrote:

Problem is that the next if statement will be useless, since R = N the statement R < N will always be false.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/baruchbarzel/NatureBiotech-31-720/issues/1#issuecomment-340680766, or mute the thread https://github.com/notifications/unsubscribe-auth/ALDaHGM_KPvuk7vk9Qof4wKCvWKGJYKNks5sxsq0gaJpZM4QMPK8 .