Generalization in the XCSF classifier system: analysis, improvement, and extension

We analyze generalization in XCSF and introduce three improvements. We begin by showing that the types of generalizations evolved by XCSF can be influenced by the input range. To explain these results we present a theoretical analysis of the convergence of classifier weights in XCSF which highlights a broader issue. In XCSF, because of the mathematical properties of the Widrow-Hoff update, the convergence of classifier weights in a given subspace can be slow when the spread of the eigenvalues of the autocorrelation matrix associated with each classifier is large. As a major consequence, the system's accuracy pressure may act before classifier weights are adequately updated, so that XCSF may evolve piecewise constant approximations, instead of the intended, and more efficient, piecewise linear ones. We propose three different ways to update classifier weights in XCSF so as to increase the generalization capabilities of XCSF: one based on a condition-based normalization of the inputs, one based on linear least squares, and one based on the recursive version of linear least squares. Through a series of experiments we show that while all three approaches significantly improve XCSF, least squares approaches appear to be best performing and most robust. Finally we show how XCSF can be extended to include polynomial approximations.     

Monomers of human beta 1 beta 1 alcohol dehydrogenase exhibit activity that differs from the dimer

A previously unreported enzymatic activity is described for monomers of the beta 1 beta 1 isoenzyme of human alcohol dehydrogenase that were prepared from dimeric enzyme by freeze-thaw in liquid nitrogen. Whereas the dimeric enzyme has optimal activity at low substrate concentrations (2.5 mM ethanol, 50 microM NAD+; "low Km" activity), the monomer has its highest activity at high substrate concentrations (1.5 M ethanol, 2.5 mM NAD+; "high Km" activity). While the activity of the monomer does not appear to be saturated at 1.5 M ethanol, its maximal activity at this high ethanol concentration exceeds the Vmax of the dimer by about 3-fold. The apparent Km of NAD+ with monomers is 270 microM, and no activity could be detected with nicotinamide mononucleotide as cofactor. During gel filtration the high Km activity elutes at a lower apparent molecular weight position than the dimer. The kinetics of monomer-to-dimer reassociation are consistent with a second-order process with a rate constant of 240 M-1 s-1. The reassociation rate is markedly enhanced by the presence of NAD+. During refolding of beta 1 beta 1 following denaturation in 6 M guanidine hydrochloride, an enzyme species with high Km activity and spectral properties similar to the freeze-thaw monomer is observed, indicating that a catalytically active monomer is an intermediate in the refolding pathway. The enzymatic activity of the monomer implies that the intersubunit contacts of beta 1 beta 1 are not crucial in establishing a catalytically competent enzyme. However, the differences in specific activity and Km between monomer and dimer suggest that dimerization may serve to modulate the catalytic properties.     

The effect of physician-controlled health insurance. U.S. v. Oregon State Medical Society

The trial record in an antitrust case against the Oregon State Medical Society, finally decided in 1952, was examined to reconstruct the behavior of a competitive market for health insurance coverage. Health insurers, called "hospital associations," were found to have engaged individually in cost-control efforts similar to, but possibly more aggressive than, today's utilization review under professional sponsorship. The subsequent disappearance of these insurer-initiated cost controls in Oregon is traced to the medical society's organization of a competing Blue Shield plan as a model of insurer conduct and to a simultaneous boycott by physicians of the hospital associations as long as they persisted in questioning doctors' practices. Some modern parallels are noted, and the advantages of fostering privately sponsored cost-control efforts are suggested.     

The complexity of approximately counting stable roommate assignments

We investigate the complexity of approximately counting stable roommate assignments in two models: (i) the k-attribute model, in which the preference lists are determined by dot products of “preference vectors” with “attribute vectors” and (ii) the k-Euclidean model, in which the preference lists are determined by the closeness of the “positions” of the people to their “preferred positions”. Exactly counting the number of assignments is #P-complete, since Irving and Leather demonstrated #P-completeness for the special case of the stable marriage problem (Irving and Leather, 1986 [11]). We show that counting the number of stable roommate assignments in the k-attribute model (#k-attribute SR, $k?4$) and the 3-Euclidean model (#k-Euclidean SR, $k?3$) is interreducible, in an approximation-preserving sense, with counting independent sets (of all sizes) ($\mathrm{\#}\mathit{IS}$) in a graph, or counting the number of satisfying assignments of a Boolean formula ($\mathrm{\#}\mathit{SAT}$). This means that there can be no FPRAS for any of these problems unless NP = RP. As a consequence, we infer that there is no FPRAS for counting stable roommate assignments ($\mathrm{\#}\mathit{SR}$) unless NP = RP. Utilizing previous results by Chebolu, Goldberg and Martin (2010) [3], we give an approximation-preserving reduction from counting the number of independent sets in a bipartite graph ($\mathrm{\#}\mathit{BIS}$) to counting the number of stable roommate assignments both in the 3-attribute model and in the 2-Euclidean model. $\mathrm{\#}\mathit{BIS}$ is complete with respect to approximation-preserving reductions in the logically-defined complexity class $\mathrm{\#}\mathrm{RH}{\mathrm{\Pi }}_1$. Hence, our result shows that an FPRAS for counting stable roommate assignments in the 3-attribute model would give an FPRAS for all $\mathrm{\#}\mathrm{RH}{\mathrm{\Pi }}_1$. We also show that the 1-attribute stable roommate problem always has either one or two stable roommate assignments, so the number of assignments can be determined exactly in polynomial time.     

Optical sensing of biomolecules using microring resonators

A biosensor application of vertically coupled glass microring resonators with Q/spl sim/12 000 is introduced. Using balanced photodetection, very high signal to noise ratios, and thus high sensitivity to refractive index changes (limit of detection of 1.8/spl times/10/sup -5/ refractive index units), are achieved. Ellipsometry and X-ray photoelectron spectroscopy results indicate successful modification of biosensor surfaces. Experimental data obtained separately for a bulk change of refractive index of the medium and for avidin-biotin binding on the ring surface are reported. Excellent repeatability and close-to-complete surface regeneration after binding are experimentally demonstrated.     

SNR estimation in time-varying fading channels

Signal-to-noise ratio (SNR) estimation is considered for phase-shift keying communication systems in time-varying fading channels. Both data-aided (DA) estimation and nondata-aided (NDA) estimation are addressed. The time-varying fading channel is modeled as a polynomial-in-time. Inherent estimation accuracy limitations are examined via the Cramer-Rao lower bound, where it is shown that the effect of the channel's time variation on SNR estimation is negligible. A novel maximum-likelihood (ML) SNR estimator is derived for the time-varying channel model. In DA scenarios, where the estimator has a simple closed-form solution, the exact performance is evaluated both with correct and incorrect (i.e., mismatched) polynomial order. In NDA estimation, the unknown data symbols are modeled as random, and the marginal likelihood is used. The expectation-maximization algorithm is proposed to iteratively maximize this likelihood function. Simulation results show that the resulting estimator offers statistical efficiency over a wider range of scenarios than previously published methods.     

ATSC DTV Receiver Implementation

Receivers of broadcast digital television (DTV) service operate in an often difficult environment of electrical interference, multiple other TV signals in close frequency proximity, multipath, wide dynamic range input signals, and uncertain antenna choice and installation. Receivers must demodulate and decode the signal and optimize its processing for different display technologies-a process that can include format conversion between progressive and interlaced scanning and different screen pixel counts. Data that supports the new services enabled by digital transmission must be decoded and processed. Receiver designers must implement the required functions cost-effectively even as they strive to provide performance and feature differentiation from their competitors' products. This paper describes all of the DTV receiver functions and references the associated standards. Emphasis is given to the difficult areas of signal reception and demodulation and to sections of the receiver that enable attractive and recognizable consumer features.     

Maximizing Reward in a Non-Stationary Mobile Robot Environment

The ability of a robot to improve its performance on a task can be critical, especially in poorly known and non-stationary environments where the best action or strategy is dependent upon the current state of the environment. In such systems, a good estimate of the current state of the environment is key to establishing high performance, however quantified. In this paper, we present an approach to state estimation in poorly known and non-stationary mobile robot environments, focusing on its application to a mine collection scenario, where performance is quantified using reward maximization. The approach is based on the use of augmented Markov models (AMMs), a sub-class of semi-Markov processes. We have developed an algorithm for incrementally constructing arbitrary-order AMMs on-line. It is used to capture the interaction dynamics between a robot and its environment in terms of behavior sequences executed during the performance of a task. For the purposes of reward maximization in a non-stationary environment, multiple AMMs monitor events at different timescales and provide statistics used to select the AMM likely to have a good estimate of the environmental state. AMMs with redundant or outdated information are discarded, while attempting to maintain sufficient data to reduce conformation to noise. This approach has been successfully implemented on a mobile robot performing a mine collection task. In the context of this task, we first present experimental results validating our reward maximization performance criterion. We then incorporate our algorithm for state estimation using multiple AMMs, allowing the robot to select appropriate actions based on the estimated state of the environment. The approach is tested first with a physical robot, in a non-stationary environment with an abrupt change, then with a simulation, in a gradually shifting environment.     

Analysis of flip-chip packaging challenges on copper/low-k interconnects

An interfacial-fracture-mechanics-based simulation methodology has been developed to study the flip-chip packaging effect on the copper/low-k structures. Multilevel submodeling techniques have been used to bridge the scale difference between the flip-chip packages and the metal/dielectric stacks. To achieve a smaller feature size and higher speed in future chips, SiO/sub 2/ can be replaced with low-k dielectric material in all via and trench layers or the number of metal layers can be increased. The effect of both packaging options has been evaluated. With either option, the future flip-chip copper/low-k packages are facing higher possibilities of adhesive or cohesive failure near the low-k interface. This paper provides a quantitative evaluation of the increased risk, thus providing guidelines to the next level of low-k flip-chip packages.