Phase transition in the spiked random tensor with Rademacher prior

TL;DR

This paper establishes the exact critical threshold for distinguishing spiked from unspiked random tensors with Rademacher prior, linking tensor detection to phase transitions in the Ising p-spin spin glass model.

Contribution

It proves that the previously identified threshold $eta_p$ precisely separates the regimes of distinguishability and indistinguishability, using analysis of the high temperature behavior of the Ising p-spin model.

Findings

The critical threshold $eta_p$ is the exact phase transition point.

Distinguishability is possible above $eta_p$, impossible below.

The threshold corresponds to a phase transition in the Ising p-spin model.

Abstract

We consider the problem of detecting a deformation from a symmetric Gaussian random $p$-tensor $(p\geq 3)$ with a rank-one spike sampled from the Rademacher prior. Recently in Lesieur et al. (2017), it was proved that there exists a critical threshold $\beta_p$ so that when the signal-to-noise ratio exceeds $\beta_p$, one can distinguish the spiked and unspiked tensors and weakly recover the prior via the minimal mean-square-error method. On the other side, Perry, Wein, and Bandeira (2017) proved that there exists a $\beta_p'<\beta_p$ such that any statistical hypothesis test can not distinguish these two tensors, in the sense that their total variation distance asymptotically vanishes, when the signa-to-noise ratio is less than $\beta_p'$. In this work, we show that $\beta_p$ is indeed the critical threshold that strictly separates the distinguishability and indistinguishability between the two tensors under the total variation distance. Our approach is based on a subtle analysis of the high temperature behavior of the pure $p$-spin model with Ising spin, arising initially from the field of spin glasses. In particular, we identify the signal-to-noise criticality $\beta_p$ as the critical temperature, distinguishing the high and low temperature behavior, of the Ising pure $p$-spin mean-field spin glass model.