We introduce a novel multivariate Laplace (MVL) distribution as a sparsity promoting prior for Bayesian source localization that allows the specification of constraints between and within sources. We represent the MVL distribution as a scale mixture that induces a coupling between source variances instead of their means. Approximation of the posterior marginals using expectation propagation is ...

A lone pilot assisted CDMA system is capable of separating the interference components included in a received input by solving a simultaneous equation. This real performance, however, depends on the property of the lone pilot response matrix which is given by the user spreading sequences and the multipath conditions. Consequently, it does not work well whenever the matrix has a wrong property. ...

The two well known learning algorithms of recurrent neural networks are the back-propagation (Rumelhart & el al., Werbos) and the forward propagation (Williams and Zipser). The main drawback of back-propagation is its off-line backward path in time for error cumulation. This violates the on-line requirement in many practical applications. Although the forward propagation algorithm can be used i...

We investigate how insights from statistical physics, namely survey propagation, can improve decoding of a particular class of sparse error correcting codes. We show that a recently proposed algorithm, time averaged belief propagation, is in fact intimately linked to a specific survey propagation for which Parisi's replica symmetry breaking parameter is set to zero, and that the latter is alway...

Iterative aggregation-disaggregation methods for numerical computing of stationary probability distribution vectors of stochastic matrices are studied. The methods can use arbitrary numbers of levels and of smoothing steps. A formula for the error propagation is derived. Using this formula, some asymptotic convergence properties of these methods for non-symmetric problems are demonstrated.

This paper discusses the usage of sparse sets for integer domain implementation over traditional representations. A first benefit of sparse sets is that they are very cheap to trail and restore. A second key advantage introduced in this work is that sparse sets permit to get delta changes with a very limited cost, allowing efficient incremental propagation. Sparse sets can also be used to repre...

A new maximum a posteriori (MAP) equalizer is proposed for digital radio links affected by large multipath delays. The “sparse” nature of the channel, where a few nonzero powerful taps are spaced by many negligible taps, is exploited to achieve a complexity proportional to the number of nonzero taps. When the channel is time-varying, an efficient nonlinear Kalman-like channel estimator is emplo...

the matrix functions appear in several applications in engineering and sciences. the computation of these functions almost involved complicated theory. thus, improving the concept theoretically seems unavoidable to obtain some new relations and algorithms for evaluating these functions. the aim of this paper is proposing some new reciprocal for the function of block anti diagonal matrices. more...

Microdamage resulting from fatigue or 'wear and tear' loading contributes to bone fragility; however, the full extent of its influence is not completely understood. Linear microcracks (∼50-100 μm) and diffuse damage (clusters of sublamellar-sized cracks) are the two major bone microdamage types, each with different mechanical and biological consequences. Healthy bone, due to its numerous micros...

The chemical nature and functional significance of mitochondrial flashes associated with fluctuations in mitochondrial membrane potential is unclear. Using a ratiometric pH probe insensitive to superoxide, we show that flashes reflect matrix alkalinization transients of ∼0.4 pH units that persist in cells permeabilized in ion-free solutions and can be evoked by imposed mitochondrial depolarizat...