Sunho Lee

In this paper, we consider a proactive eavesdropping scenario, where a central monitor tries to intercept the information exchanged between a pair of suspicious entities through amplify-and-forward full-duplex relays and a cooperative jammer. Specifically, the eavesdropping relays simultaneously listen to the suspicious transmitter and forward the eavesdropped information to the central monitor. At the same time, the jammer broadcasts jamming signals to maintain the data rate of the suspicious users below the channel capacity of the central monitor so that the eavesdropped information can be successfully decoded by the central monitor. In this system, we jointly design the receive combining vector at the central monitor, the relay precoders at the eavesdropping relays and the transmit covariance matrix at the jammer to maximize the eavesdropping rate. First, we examine the case of a single eavesdropping relay equipped with a single antenna and provide some useful insights. Also, an effective two-layer optimization method is proposed to obtain the globally optimal solution. Then, we study a general case of multiple eavesdropping relays with multiple antennas, and solve a semi-definite relaxation problem. Numerical results verify the efficiency of our proposed solutions for both single and multiple relay cases and show considerable performance gains over conventional schemes.

In this letter, we propose an efficient algorithm which reduces the complexity of conventional vector perturbation schemes by searching the real and imaginary components of a perturbation vector individually. To minimize a performance loss induced from the decoupled joint search, we apply diagonal precoding at the transmitter whose parameters are iteratively optimized to maximize the chordal distance between subspaces spanned by the real and imaginary components. We also propose a simple non-iterative method with a slight performance loss which can achieve a significant complexity reduction compared to the conventional vector perturbation schemes.

In this letter, we present an improved precoding technique which reduces a modulo loss in vector perturbation with low complexity. Instead of searching perturbation vectors in the infinite lattice, the proposed scheme restricts the search range by utilizing the distribution of the perturbation vector depending on transmitted data. As a result, we can achieve significant complexity savings at the transmitter while providing better performance compared to the original vector perturbation.

A major interest in gene expression microarray studies is to develop an accurate classifier which can be adopted in clinical practice. The usage of large numbers of genes with small data samples may lead to overfitting in classification, and generate promising, but often nonreproducible results. Therefore, assessing the reproducibility of a classifier is necessary. Appropriate methods for validating a developed classifier and estimating its predicting accuracy are discussed. In addition, some mistakes that can arise in the cross validation process are reviewed using published articles in prominent medical journals, to prevent the indefinite results of a classifier development from leading to inappropriate treatment.

Poisson models can be widely used for analyzing count data and belong to some broad classes of discrete distributions such as a Katz family, a double exponential family and a family of discrete distributions with a finite first moment. Several procedures for testing goodness of fit of a Poisson model against these classes are discussed and simulations are conducted to compare their performance for detecting departures from Poisson models.