A spin glass model for reconstructing nonlinearly encrypted signals corrupted by noise

An encryption of a signal ${\bf s}\in\mathbb{R^N}$ is a random mapping ${\bf s}\mapsto \textbf{y}=(y_1,\ldots,y_M)^T\in \mathbb{R}^M$ which can be corrupted by an additive noise. Given the Encryption Redundancy Parameter (ERP) $\mu=M/N\ge 1$, the signal strength parameter $R=\sqrt{\sum_i s_i^2/N}$, and the ('bare') noise-to-signal ratio (NSR) $\gamma\ge 0$, we consider the problem of reconstructing ${\bf s}$ from its corrupted image by a Least Square Scheme for a certain class of random Gaussian mappings. The problem is equivalent to finding the configuration of minimal energy in a certain version of spherical spin glass model, with squared Gaussian-distributed random potential. We use the Parisi replica symmetry breaking scheme to evaluate the mean overlap $p_{\infty}\in [0,1]$ between the original signal and its recovered image (known as 'estimator') as $N\to \infty$, which is a measure of the quality of the signal reconstruction. We explicitly analyze the general case of linear-quadratic family of random mappings and discuss the full $p_{\infty} (\gamma)$ curve. When nonlinearity exceeds a certain threshold but redundancy is not yet too big, the replica symmetric solution is necessarily broken in some interval of NSR. We show that encryptions with a nonvanishing linear component permit reconstructions with $p_{\infty}>0$ for any $\mu>1$ and any $\gamma<\infty$, with $p_{\infty}\sim \gamma^{-1/2}$ as $\gamma\to \infty$. In contrast, for the case of purely quadratic nonlinearity, for any ERP $\mu>1$ there exists a threshold NSR value $\gamma_c(\mu)$ such that $p_{\infty}=0$ for $\gamma>\gamma_c(\mu)$ making the reconstruction impossible. The behaviour close to the threshold is given by $p_{\infty}\sim (\gamma_c-\gamma)^{3/4}$ and is controlled by the replica symmetry breaking mechanism.