The p-norm generalization of the LMS algorithm for adaptive filtering

Recently much work has been done analyzing online machine learning algorithms in a worst case setting, where no probabilistic assumptions are made about the data. This is analogous to the H/sup /spl infin// setting used in adaptive linear filtering. Bregman divergences have become a standard tool for analyzing online machine learning algorithms. Using these divergences, we motivate a generalization of the least mean squared (LMS) algorithm. The loss bounds for these so-called p-norm algorithms involve other norms than the standard 2-norm. The bounds can be significantly better if a large proportion of the input variables are irrelevant, i.e., if the weight vector we are trying to learn is sparse. We also prove results for nonstationary targets. We only know how to apply kernel methods to the standard LMS algorithm (i.e., p=2). However, even in the general p-norm case, we can handle generalized linear models where the output of the system is a linear function combined with a nonlinear transfer function (e.g., the logistic sigmoid).     

Empirical characterization of wideband indoor radio channel at 5.3GHz

Characteristics of wideband indoor radio channel at 5.3 GHz were defined based on an extensive measurement campaign using a wideband channel sounder with 19 ns delay resolution. Pathloss exponents were 1.3-1.5 in LOS and 2.9-4.8 in non-line of sight (NLOS). Large difference in NLOS exponents was due to different dominating propagation mechanisms in different types of building structures. The delay dispersion was characterized by cumulative distribution functions (CDF) of the RMS delay spreads, the values for CDF=0.9 varied from 20 to 180 ns in different setups in an office building and large hall environments. The correlation functions of the radio channel in spatial and frequency domains were extracted. Small scale models for five typical indoor scenarios were developed using tapped delay lines     

Polarization Behaviours at 2, 5 and 60 GHz for Indoor Mobile Communications

The behaviours of linear polarizations at 2.15, 5.3 and 61.7 GHz in corridors are studied in this paper. It shows that there is no significant difference between the received powers for vertical and horizontal polarizations. Depolarization is obvious at 2.15 GHz due to different antenna type is applied at the receiver, and it is more serious in non-line-of-sight (NLOS) cases.     

Potential of amorphous Mo–Si–N films for nanoelectronic applications

The properties of amorphous metallic molybdenum–silicon–nitrogen (Mo–Si–N) films were characterised for use in nanoelectronic applications. The films were deposited by co-sputtering of molybdenum and silicon targets in a gas mixture of argon and nitrogen. The atomic composition, microstructure and surface roughness were studied by RBS, TEM and AFM analyses, respectively. The electrical properties were investigated in the temperature range 80 mK to 300 K. No transition into a superconductive state was observed. Nanoscale wires were fabricated using electron beam lithography with their properties measured as a function of temperature.     

Online learning with kernels

Kernel-based algorithms such as support vector machines have achieved considerable success in various problems in batch setting, where all of the training data is available in advance. Support vector machines combine the so-called kernel trick with the large margin idea. There has been little use of these methods in an online setting suitable for real-time applications. In this paper, we consider online learning in a reproducing kernel Hilbert space. By considering classical stochastic gradient descent within a feature space and the use of some straightforward tricks, we develop simple and computationally efficient algorithms for a wide range of problems such as classification, regression, and novelty detection. In addition to allowing the exploitation of the kernel trick in an online setting, we examine the value of large margins for classification in the online setting with a drifting target. We derive worst-case loss bounds, and moreover, we show the convergence of the hypothesis to the minimizer of the regularized risk functional. We present some experimental results that support the theory as well as illustrating the power of the new algorithms for online novelty detection.     

Polarized infrared spectra: Part 1. Stretched polymer method. Fundamental vibrations of some mono- and di-substituted benzenes

Polarized IR spectra of five benzene derivatives (four of C_2v symmetry and one of D_2h symmetry) have been recorded using the stretched polymer method. It has been found that the values of the dichroic ratios depend systematically on the symmetry class of the vibration in question. The IR-active vibrations can thus be unambiguously divided into the correct symmetry classes. The general validity of the method, as well as its value for vibrational analysis, is discussed.     

Comparison of MIMO antenna configurations in picocell and microcell environments

This paper presents the results achieved with a dual-polarized multiple-input multiple-output (MIMO) measurement system in the 2 GHz range. Results from continuous measurement routes were used in evaluating and comparing different MIMO antenna configurations. Different pattern and polarization diversity possibilities were studied using two methods: elements were selected from the antenna arrays used in measurements, and as another option, in the mobile station the incident waves were estimated and used in different dipole antenna arrays. The capacity limit seems to be higher in an indoor picocell than in an outdoor microcell environment. At the mobile station, directive elements result in 35% higher average capacities than those of the omnidirectional elements; however, the capacity of the directive elements also depends on the azimuth direction of arrival of the incident field. Dual-polarized antenna configurations have approximately 14% higher capacities than copolarized configurations. Increasing the number of mobile antenna elements increases the capacity in those environments where the angular spread of the incident field is large. Increasing the distance between elements at the fixed station increases the capacity - especially in microcells where signals arrive from specific directions.     

Sequential prediction of individual sequences under general lossfunctions

We consider adaptive sequential prediction of arbitrary binary sequences when the performance is evaluated using a general loss function. The goal is to predict on each individual sequence nearly as well as the best prediction strategy in a given comparison class of (possibly adaptive) prediction strategies, called experts. By using a general loss function, we generalize previous work on universal prediction, forecasting, and data compression. However, here we restrict ourselves to the case when the comparison class is finite. For a given sequence, we define the regret as the total loss on the entire sequence suffered by the adaptive sequential predictor, minus the total loss suffered by the predictor in the comparison class that performs best on that particular sequence. We show that for a large class of loss functions, the minimax regret is either &thetas;(log N) or Ω(√ℒlog N), depending on the loss function, where N is the number of predictors in the comparison class andℒ is the length of the sequence to be predicted. The former case was shown previously by Vovk (1990); we give a simplified analysis with an explicit closed form for the constant in the minimax regret formula, and give a probabilistic argument that shows this constant is the best possible. Some weak regularity conditions are imposed on the loss function in obtaining these results. We also extend our analysis to the case of predicting arbitrary sequences that take real values in the interval [0,1]