Cancer classification based on microarray data has gained attention in recent years from the bioinformatics community, due to a high death toll of cancer and the significance of early diagnosis. Among the many steps in cancer classification, one that is underexplored and can significantly affect the classification performance is data transformation. We develop two transformation techniques, called the unity-based normalization with min-max interval (UBMI) and the standard score with trimmed mean (SSTM), and compare them with existing techniques in terms of accuracy, sensitivity and specificity. The results show that our proposed methods outperform the methods tested. Overall, for example, the SSTM achieves the highest values of accuracy, sensitivity, and specificity in 73 out of 138 cases. The UBMI is the runner up, with 51 winning cases. This advantage confirms the ability of the UBMI and SSTM to accentuate the difference between samples of distinct classes, and highlights the importance of data transformation, a step that otherwise is usually overlooked.

Received signal strength (RSS)-based localization has been widely used in location-aware applications due to its low cost and low complexity. The accuracy of RSS-based localization depends on the values of parameters used in the path-loss model, which is specific to the operating environment. Given the dependence, however, most work assumes the path-loss parameters are fixed and known, an assumption that costs positioning accuracy. In this paper, we estimate the target position and the path-loss parameters jointly for an arbitrary distribution of noise. We formulate the estimation problem as an optimization that has a prominent geometric structure. To achieve any level of positioning accuracy, as measured by the outage on the distance error, we prove that the anchors must be placed in certain ways and that the sample size of RSSs must be large enough. The method to place anchors and a guaranteed sample size are given in the paper. The proposed framework has practical utility to localization in unknown environments when the target node must be simple and yet able to supply accurate positioning.

First passage time problems arise in many fields, but few of them are explicitly solved. Motivated by an application to synchronization of clocks, this paper obtains the probability distribution for the first passage time for a Brownian motion with quadratic drift to exit from two constant barriers that are not necessarily symmetric. The expression for the distribution is explicit, consisting of elementary functions and functions that are characterized by Laplace transforms. Applications of first passage times are demonstrated in the context of synchronization.

In this paper, we investigate performance of acquisition receivers employing randomized search strategies. In particular, we derive the expressions of the mean acquisition times (MATs) for the uniform randomized search strategy and the non-return-to-self uniform randomized search strategy (NRS). These two search strategies are robust against variation in the maximum dispersion of the channel, making them attractive from the standpoint of a receiver design. We also quantify the MATs of receivers that employ the uniform randomized search strategy and the multi-dwell detectors with optimal thresholds. The MATs are plotted for various search strategies over a range of signal-to-noise ratios, using a channel model that is suitable for wideband transmission. The numerical examples confirm the benefit of the uniform randomized search strategy and the NRS over the conventional serial search strategy.

Cooperative diversity techniques can improve the transmission rate and reliability of wireless networks. For systems employing such diversity techniques in slow-fading channels, outage probability and outage capacity are important performance measures. Existing studies have derived approximate expressions for these performance measures in different scenarios. In this paper, we derive the exact expressions for outage probabilities and outage capacities of three proactive cooperative diversity schemes that select a best relay from a set of relays to forward the information. The derived expressions are valid for arbitrary network topology and operating signal-to-noise ratio, and serve as a useful tool for network design.

Cooperative diversity is a useful technique to increase reliability and throughput of wireless networks. To analyze the performance gain from cooperative diversity, outage capacity is an important figure of merit that captures the inherent diversity-multiplexing trade-off in cooperative diversity schemes. In this paper we derive the outage capacity for several cooperative diversity schemes in decode-and-forward relay networks with a finite number of relay nodes. The obtained expressions are simple and applicable to arbitrary network topologies and signal-to-noise ratios in Rayleigh fading channels. The analysis shows that there exists a signal-to-noise ratio threshold, below which some cooperative diversity schemes are better than direct communication. We propose a new diversity scheme, which, compared to the conventional counterpart, offers improved performance and requires protocol overhead.

A cognitive radio can sense its environment and adapt some of its features, such as carrier frequency, transmission bandwidth, transmission power, and modulation, thus allowing dynamic reuse of the available spectrum. Due to their high degree of adaptability to environmental variations, cognitive radios are expected to utilize packet-based transmission with variable-length frames. Packet-based transmission requires the receiver to perform frame synchronization, an important enabling step that allows adaptation in cognitive radios. However, proper metrics to characterize the performance of frame synchronization for transmission of variable-length frames are currently unavailable. To address this issue, we put forth two performance metrics, namely the expected duration to complete frame synchronization and the probability of correct acquisition within a given duration. We then develop analytical expressions for these important metrics. This paper advances our understanding of frame synchronization for the continuous transmission of variable-length frames and for bursty transmission.

In this paper, we propose a search technique that takes advantage of multipath, which has long been considered deleterious for efficient communication, to aid the sequence acquisition in dense multipath channels. We consider a class of serial-search strategies and use optimization and convexity theories to determine fundamental limits of achievable mean acquisition times (MATs). In particular, we derive closed-form expressions for both the minimum and maximum MATs and the conditions for achieving these limits. We prove that a fixed-step serial search, a form of nonconsecutive serial search, achieves a near-optimal MAT. We also prove that the conventional serial search, in which consecutive cells are tested serially, should be avoided as it results in the maximum MAT. Our results are valid for all signal-to-noise ratio (SNR) values, regardless of the specifics of the detection layer and the fading distributions.

We study sequential frame synchronization based on markers (or sync words), where the received samples are observed over a window of length equal to the marker length. A decision variable is derived over the observation window. A marker is declared if the decision variable exceeds a threshold; otherwise the observation window is time-shifted by one sample. This detection strategy can be characterized by the probabilities of false alarm and of correct detection, usually represented in terms of the receiver operating characteristic (ROC). However, once the detector (and hence its ROC) is given, it is crucial from a design perspective to select an optimal operating point on the ROC. In this paper we propose a general framework to connect the ROC to the performance of frame synchronization in terms of the probability of correct acquisition in an arbitrarily long time period, valid both for constant and variable length frames. This framework can also be used to compare different marker detectors for frame synchronization.

Spread-spectrum systems with large transmission bandwidth present significant challenges from the standpoint of achieving acquisition before the communication commences. This paper investigates a rapid acquisition procedure that exploits multipath to aid the synchronization. In particular, we consider a randomized search strategy and derive the corresponding mean acquisition time (MAT). We show that the MAT of the randomized search is at most twice the MAT of the optimal serial search. Using our results, we compare various search techniques in a multipath fading channel