Thursday, April 4, 2013

1304.0942 (Aurélien Decelle et al.)

Decimation based method to improve inference using the Pseudo-Likelihood
method
   [PDF]

Aurélien Decelle, Federico Ricci-Tersenghi
In this Letter we propose a new method to infer the topology of the interaction network in a pairwise model with Ising variables. By using the pseudo-likelihood method (PLM) at high temperature, it is generally possible to distinguish between zero and non-zero couplings, because a clear gap separate the two groups. However at lower temperatures the PLM is much less effective and the result depends on subjective choices, as the value of the l1-regularizer and that of the threshold to separate non-zero couplings from null ones. We introduce a decimation procedure based on PLM, that recursively sets to zero the less significant couplings, until the variation of the pseudo-likelihood signals that relevant couplings are being removed. The new method is fully automatized and does not require any subjective choice by the user. Numerical tests shows that it performs better than standard PLM.
View original: http://arxiv.org/abs/1304.0942

No comments:

Post a Comment