Abstract
We characterize the minimum number of measurements needed to drive to zero the minimum mean squared error (MMSE) of Gaussian mixture model (GMM) input signals in the low-noise regime. The result also hints at almost phasetransition optimal recovery procedures based on a classification and reconstruction approach.

Abstract
This paper puts forth projections designs for compressive classification of Gaussian mixture models. In particular, we capitalize on the asymptotic characterization of the behavior of an (upper bound to the) misclassification probability associated with the optimal Maximum-A-Posteriori (MAP) classifier, which depends on quantities that are dual to the concepts of the diversity gain and coding gain in multi-antenna communications, to construct measurement designs that maximize the diversity-order of the measurement model. Numerical results demonstrate that the new measurement designs substantially outperform random measurements. Overall, the analysis and the designs cast geometrical insight about the mechanics of compressive classification problems.

Abstract
A transmission between two agents, Alice and Bob, over a set of parallel sub-channels is overheard by a third agent Eve, through a second set of parallel sub-channels. All sub-channels are flat with random and independent gains and additive white Gaussian noise (AWGN). Alice splits the total amount of available power among the sub-channels, with the purpose of maximizing the communication rate to Bob, under reliability and secrecy constraints. To this end, two schemes are considered. In one case the secret message is encoded with a single wiretap code and then split among the sub-channels. In the latter case the secret message is first split into a number of sub-messages, each separately encoded and transmitted on a different sub-channel. The achievable secrecy rates under a constraint on the secrecy outage probability (SOP) are derived and closed form expressions for Rayleigh fading sub-channels are obtained. In order to limit the complexity of resources optimization (power and rates) we also consider suboptimal solutions based on the selection of active sub-channels over which power is split either equally or according to a waterfilling algorithm with respect to the Alice-Bob channel.

Abstract
This paper puts forth new recovery guarantees for the source separation problem in the presence of side information, where one observes the linear superposition of two source signals plus two additional signals that are correlated with the mixed ones. By positing that the individual components of the mixed signals as well as the corresponding side information signals follow a joint Gaussian mixture model, we characterise necessary and sufficient conditions for reliable separation in the asymptotic regime of low-noise as a function of the geometry of the underlying signals and their interaction. In particular, we show that if the subspaces spanned by the innovation components of the source signals with respect to the side information signals have zero intersection, provided that we observe a certain number of measurements from the mixture, then we can reliably separate the sources, otherwise we cannot. We also provide a number of numerical results on synthetic data that validate our theoretical findings.

Abstract
This paper studies the performance associated with the classification of linear subspaces corrupted by noise with a mismatched classifier. In particular, we consider a problem where the classifier observes a noisy signal, the signal distribution conditioned on the signal class is zero-mean Gaussian with low-rank covariance matrix, and the classifier knows only the mismatched parameters in lieu of the true parameters. We derive an upper bound to the misclassification probability of the mismatched classifier and characterize its behaviour. Specifically, our characterization leads to sharp sufficient conditions that describe the absence of an error floor in the low-noise regime, and that can be expressed in terms of the principal angles and the overlap between the true and the mismatched signal subspaces. © 2015 IEEE.

Abstract
This paper considers the classification of linear subspaces with mismatched classifiers. In particular, we assume a model where one observes signals in the presence of isotropic Gaussian noise and the distribution of the signals conditioned on a given class is Gaussian with a zero mean and a low-rank covariance matrix. We also assume that the classifier knows only a mismatched version of the parameters of input distribution in lieu of the true parameters. By constructing an asymptotic low-noise expansion of an upper bound to the error probability of such a mismatched classifier, we provide sufficient conditions for reliable classification in the low-noise regime that are able to sharply predict the absence of a classification error floor. Such conditions are a function of the geometry of the true signal distribution, the geometry of the mismatched signal distributions as well as the interplay between such geometries, namely, the principal angles and the overlap between the true and the mismatched signal subspaces. Numerical results demonstrate that our conditions for reliable classification can sharply predict the behavior of a mismatched classifier both with synthetic data and in a motion segmentation and a hand-written digit classification applications. © 2016 IEEE.