A New Gradient Descent Approach for Local Learning of Fuzzy Neural Models

The majority of reported learning methods for Takagi-Sugeno-Kang fuzzy neural models to date mainly focus on the improvement of their accuracy. However, one of the key design requirements in building an interpretable fuzzy model is that each obtained rule consequent must match well with the system local behaviour when all the rules are aggregated to produce the overall system output. This is one of the distinctive characteristics from black-box models such as neural networks. Therefore, how to find a desirable set of fuzzy partitions and, hence, to identify the corresponding consequent models which can be directly explained in terms of system behaviour presents a critical step in fuzzy neural modelling. In this paper, a new learning approach considering both nonlinear parameters in the rule premises and linear parameters in the rule consequents is proposed. Unlike the conventional two-stage optimization procedure widely practised in the field where the two sets of parameters are optimized separately, the consequent parameters are transformed into a dependent set on the premise parameters, thereby enabling the introduction of a new integrated gradient descent learning approach. A new Jacobian matrix is thus proposed and efficiently computed to achieve a more accurate approximation of the cost function by using the second-order Levenberg-Marquardt optimization method. Several other interpretability issues about the fuzzy neural model are also discussed and integrated into this new learning approach. Numerical examples are presented to illustrate the resultant structure of the fuzzy neural models and the effectiveness of the proposed new algorithm, and compared with the results from some well-known methods.

Fast-electron self-collimation in a plasma density gradient

A theoretical and numerical study of fast electron transport in solid and compressed fast ignition relevant targets is presented. The principal aim of the study is to assess how localized increases in the target density (e. g., by engineering of the density profile) can enhance magnetic field generation and thus pinching of the fast electron beam through reducing the rate of temperature rise. The extent to which this might benefit fast ignition is discussed. (C) 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4729322]

Field and Laboratory Tests of Flow-Proportional Passive Samplers for Determining Average Phosphorus and Nitrogen Concentration in Rivers

Flow responsive passive samplers offer considerable potential in nutrient monitoring in catchments; bridging the gap between the intermittency of grab sampling and the high cost of automated monitoring systems. A commercially available passive sampler was evaluated in a number of river systems encapsulating a gradient in storm response, combinations of diffuse and point source pressures, and levels of phosphorus and nitrogen concentrations. Phosphorus and nitrogen are sequestered to a resin matrix in a permeable cartridge positioned in line with streamflow. A salt tracer dissolves in proportion to advective flow through the cartridge. Multiple deployments of different cartridge types were undertaken and the recovery of P and N compared with the flow-weighted mean concentration (FWMC) from high-resolution bank-side analysers at each site. Results from the passive samplers were variable and largely underestimated the FWMC derived from the bank-side analysers. Laboratory tests using ambient river samples indicated good replication of advective throughflow using pumped water, although this appeared not to be a good analogue of river conditions where flow divergence was possible. Laboratory tests also showed good nutrient retention but not elution and these issues appeared to combine to limit the utility in ambient river systems at the small catchment scale.

On the required shape corrections to the local density and generalized gradient approximations to the Kohn-Sham potentials for molecular response calculations of (hyper)polarizabilities and excitation energies

It is well known that shape corrections have to be applied to the local-density (LDA) and generalized gradient (GGA) approximations to the Kohn-Sham exchange-correlation potential in order to obtain reliable response properties in time dependent density functional theory calculations. Here we demonstrate that it is an oversimplified view that these shape corrections concern primarily the asymptotic part of the potential, and that they affect only Rydberg type transitions. The performance is assessed of two shape-corrected Kohn-Sham potentials, the gradient-regulated asymptotic connection procedure applied to the Becke-Perdew potential (BP-GRAC) and the statistical averaging of (model) orbital potentials (SAOP), versus LDA and GGA potentials, in molecular response calculations of the static average polarizability alpha, the Cauchy coefficient S-4, and the static average hyperpolarizability beta. The nature of the distortions of the LDA/GGA potentials is highlighted and it is shown that they introduce many spurious excited states at too low energy which may mix with valence excited states, resulting in wrong excited state compositions. They also lead to wrong oscillator strengths and thus to a wrong spectral structure of properties like the polarizability. LDA, Becke-Lee-Yang-Parr (BLYP), and Becke-Perdew (BP) characteristically underestimate contributions to alpha and S-4 from bound Rydberg-type states and overestimate those from the continuum. Cancellation of the errors in these contributions occasionally produces fortuitously good results. The distortions of the LDA, BLYP, and BP spectra are related to the deficiencies of the LDA/GGA potentials in both the bulk and outer molecular regions. In contrast, both SAOP and BP-GRAC potentials produce high quality polarizabilities for 21 molecules and also reliable Cauchy moments and hyperpolarizabilities for the selected molecules. The analysis for the N-2 molecule shows, that both SAOP and BP-GRAC yield reliable energies omega(i) and oscillator strengths f(i) of individual excitations, so that they reproduce well the spectral structure of alpha and S-4.(C) 2002 American Institute of Physics.

Field observation of sequestration of uranium in iron-rich sediments under sequential reduction-oxidation conditions at the DOE Oak Ridge IFRC

At the U.S. DOE Oak Ridge Integrated Field Research Challenge (ORIFRC) site, the iron content of shallow subsurface materials (i.e. weathered saprolite) is relatively high (up to 5-6% as w/w), and therefore, the forms of the iron species present plays a critical role in the long-term sequestration of uranium. A long term pilot-scale study of the bioreduction and reoxidation of uranium conducted at the ORIFRC area 3 site, adjacent to the former S-3 disposal ponds (source zone), has provided us with the opportunity to study the impact of iron species on the sequestration of U(VI). The aqueous U(VI) concentrations at the site were decreased to below the EPA MCL through the intermittent injection of ethanol as the electron donor. Previous field tests indicated that both oxygen and nitrate could oxidize the bioreduced U(IV) and cause a short-term rebound of aqueous phase uranium concentration after the oxidative agents were delivered directly to the bioreduced zone.

A field test has been conducted to examine the long-term effect of exposure of bioreduced sediments to nitrate in contaminated groundwater for more than 1,380 days at the Area 3 site. Contaminated groundwater was allowed to invade the previously bioreduced zone via the natural groundwater gradient after an extended period in which reducing conditions were maintained and the bioreduced zone was protected from the influx of upgradient contaminated groundwater. The geochemical response to the invasion of contaminated groundwater was dependent on whether the monitoring location is in the middle or the fringe of the previously bioreduced zone. In general, the nitrate concentrations in the previously bioreduced area, increased gradually from near zero to ~50-300 mM within 200 days and then stabilized. The pH declined from bioreduced levels of 6.2-6.7 to below 5.0. Uranium concentrations rebounded in all monitoring wells but at different rates. At most locations U concentrations rebounded, declined and then rebounded again. Methane gas disappeared while a significant level (20,000 to 44,000 ppmv) N2O was found in the groundwater of monitoring wells after three years of reoxidization.

The U(IV) in sediments was mainly reoxidized to U(VI) species. Based on XANES analysis, the predominate uranium in all samples after re-oxidation was similar to a uranyl nitrate form. But the U content in the sediment remained as high as that determined after bioreduction activates were completed, indicating that much of the U is still sequestrated in situ. SEM observations of surged fine sediments revealed that clusters of colloidal-sized (200-500nm) U-containing precipitates appeared to have formed in situ, regardless from sample of FW106 in non-bioactivity control area or of pre-bioreduced FW101-2 and FW102-3. Additionally, SEM-EDS and microprobe analysis, showed that the U-containing precipitates (~1% U) in FW106 are notably higher in Fe, compared to the precipitates (~1-2.5% U) from FW101-2 and FW102-3. However, XRF analysis indicated that the U content was remained as high as 2180 and 1810 mg/kg with U/Fe ratio at 0.077 and 0.055 vs 0.037 g/g, respectively in pre-bioreduced FW101-2 and FW102-3, suggesting more U sequestrated by Fe in pre-bioreduced sediments.

Effects of resistive magnetic field on fast electron divergence measured in experiments

Transport of fast electrons driven by an ultraintense laser through a tracer layer buried in solid targets is studied by particle-in-cell simulations. It is found that intense resistive magnetic fields, having a magnitude of several thousand Tesla, are generated at the interfaces of the materials due to the steep resistivity gradient between the target and tracer layer. Such magnetic fields can significantly inhibit the fast electron propagation. The electrons that can penetrate the first interface are mostly confined in the buried layer by the magnetic fields and cause heating of the tracer layer. The lateral extent of the heated region can be significantly larger than that of the relativistic electron beam. This finding suggests that the relativistic electron divergence inferred from Ká x-ray emission in experiments might be overestimated.

Snapshot gradient-recalled echo-planar images of rat brains at long echo time at 9.4 T.

With improved B 0 homogeneity along with satisfactory gradient performance at high magnetic fields, snapshot gradient-recalled echo-planar imaging (GRE-EPI) would perform at long echo times (TEs) on the order of T2*, which intrinsically allows obtaining strongly T2*-weighted images with embedded substantial anatomical details in ultrashort time. The aim of this study was to investigate the feasibility and quality of long TE snapshot GRE-EPI images of rat brain at 9.4 T. When compensating for B 0 inhomogeneities, especially second-order shim terms, a 200 x 200 microm2 in-plane resolution image was reproducibly obtained at long TE (>25 ms). The resulting coronal images at 30 ms had diminished geometric distortions and, thus, embedded substantial anatomical details. Concurrently with the very consistent stability, such GRE-EPI images should permit to resolve functional data not only with high specificity but also with substantial anatomical details, therefore allowing coregistration of the acquired functional data on the same image data set.

Male reproductive competition in the field crickets : gryllus veletis and G. pennsylvanicus

Sexual behavior in the field crickets, Gryllus veletis and G. pennsylvanicus , was studied in outdoor arenas (12 m2) at high and low levels of population density in 1983 and 1984. Crickets were weighed, individually marked, and observed from 2200 until 0800 hrs for at least 9 continuous nights. Calling was measured at 5 min intervals, and movement and matings were recorded hourly. Continuous 24 hr observations were also conducted,·and occurrences of aggressive and courtship songs were noted. The timing of males searching, calling, courting, and fighting for females should coincide with female movement and mating patterns. For most samples female movement and matings occurred at night in the 24 hr observations and were randomly distributed with time for both species in the 10 hr observations. Male movement for G. veletis high density only was enhanced at night in the 24 hr observations, however, males called more at night in both species at high and low densities. Male movement was randomly distributed with time in the 10 hr observations, and calling increased at dawn for the G. pennsylvanicus 1984 high density sample, but was randomly distributed in other samples. Most courtship and aggression songs in the 24 hr observations were too infrequent for statistical testing and generally did not coincide with matings. Assuming residual reproductive value, and costs attached to a male trait in terms of future reproductive success decline with age, males should behave in more costly ways with age; by calling and moving more with age. Consequently, mating rates should increase with age. Female behavior may not change with age. G. veletis , females moved more with age at both low density samples, however, crickets moved less with age at high density. G. pennsylvanicus females moved more with age in the 1984 low density sample, whereas crickets moved less with age in the 1983 high density sample. For both species males in the 1984 high density samples called less with age. For G. pennsylvanicus in 1983 calling and mating rates increased with age. Mating rates decreased with age for G. veletis males in the high density sample. Aging may not affect cricket behavior. As population density increases fewer calling sites become available, costs of territoriality increase, and matings resulting from non-calling behavior should increase. For both species the amount of calling and in G. veletis the distance travelled per night was not different between densities. G. pennsylvanicus males and females moved more at low density. At the same deneity levels there were no differences in calling, mating, and, movement rates in G. veletis , however, G. pennsylvanicus males moved more at high density in 1983 than 1984. There was a positive relationship between calling and mating for the G. pennsylvanicus low density sample only, and selection was acting directly to increase calling. For both species no relationships between movement and mating success was found, however, the selection gradient on movement in the G. veletis high density population was significant. The intensity of selection was not significant and was probably due to the inverse relationship between displacement and weight. Larger males should call more, mate more, and move less than smaller males. There were no correlations between calling and individual weight, and an inverse correlation between movement and size in the G. veletis high density population only. In G. pennsylvanicus , there was a positive correlation between individual weight and mating, but, some correlate of weight was under counter selection pressure and-prevented significance of the intensity of selection. In contrast, there was an inverse correlation in the G.·veletis low density B sample. Both measures of selection intensities were significant and showed that weight only was under selection pressures. An inverse correlation between calling and movement was found for G. veletis at low density only. Because males are territorial, females are predicted to move more than males, however, if movement is a mode of male-male reproductive competition then males may move more than females. G. pennsylvanicus males moved more than females in all samples, however, G. veletis males and females moved similar distances at all densities. The variation in relative mating success explained by calling scores, movement, and weight for both species and all samples were not significant In addition, for both species and all samples the intensity of selection never equalled the opportunity for selection.

Policy gradient based Reinforcement Learning for real autonomous underwater cable tracking

This paper proposes a field application of a high-level reinforcement learning (RL) control system for solving the action selection problem of an autonomous robot in cable tracking task. The learning system is characterized by using a direct policy search method for learning the internal state/action mapping. Policy only algorithms may suffer from long convergence times when dealing with real robotics. In order to speed up the process, the learning phase has been carried out in a simulated environment and, in a second step, the policy has been transferred and tested successfully on a real robot. Future steps plan to continue the learning process on-line while on the real robot while performing the mentioned task. We demonstrate its feasibility with real experiments on the underwater robot ICTINEU AUV

Gradient-based reinforcement learning techniques for underwater robotics behavior learning

Darrerament, l'interès pel desenvolupament d'aplicacions amb robots submarins autònoms (AUV) ha crescut de forma considerable. Els AUVs són atractius gràcies al seu tamany i el fet que no necessiten un operador humà per pilotar-los. Tot i això, és impossible comparar, en termes d'eficiència i flexibilitat, l'habilitat d'un pilot humà amb les escasses capacitats operatives que ofereixen els AUVs actuals. L'utilització de AUVs per cobrir grans àrees implica resoldre problemes complexos, especialment si es desitja que el nostre robot reaccioni en temps real a canvis sobtats en les condicions de treball. Per aquestes raons, el desenvolupament de sistemes de control autònom amb l'objectiu de millorar aquestes capacitats ha esdevingut una prioritat. Aquesta tesi tracta sobre el problema de la presa de decisions utilizant AUVs. El treball presentat es centra en l'estudi, disseny i aplicació de comportaments per a AUVs utilitzant tècniques d'aprenentatge per reforç (RL). La contribució principal d'aquesta tesi consisteix en l'aplicació de diverses tècniques de RL per tal de millorar l'autonomia dels robots submarins, amb l'objectiu final de demostrar la viabilitat d'aquests algoritmes per aprendre tasques submarines autònomes en temps real. En RL, el robot intenta maximitzar un reforç escalar obtingut com a conseqüència de la seva interacció amb l'entorn. L'objectiu és trobar una política òptima que relaciona tots els estats possibles amb les accions a executar per a cada estat que maximitzen la suma de reforços totals. Així, aquesta tesi investiga principalment dues tipologies d'algoritmes basats en RL: mètodes basats en funcions de valor (VF) i mètodes basats en el gradient (PG). Els resultats experimentals finals mostren el robot submarí Ictineu en una tasca autònoma real de seguiment de cables submarins. Per portar-la a terme, s'ha dissenyat un algoritme anomenat mètode d'Actor i Crític (AC), fruit de la fusió de mètodes VF amb tècniques de PG.

Suprathermal electron evolution in a Parker spiral magnetic field

Suprathermal electrons (>70 eV) form a small fraction of the total solar wind electron density but serve as valuable tracers of heliospheric magnetic field topology. Their usefulness as tracers of magnetic loops with both feet rooted on the Sun, however, most likely fades as the loops expand beyond some distance owing to scattering. As a first step toward quantifying that distance, we construct an observationally constrained model for the evolution of the suprathermal electron pitch-angle distributions on open field lines. We begin with a near-Sun isotropic distribution moving antisunward along a Parker spiral magnetic field while conserving magnetic moment, resulting in a field-aligned strahl within a few solar radii. Past this point, the distribution undergoes little evolution with heliocentric distance. We then add constant (with heliocentric distance, energy, and pitch angle) ad-hoc pitch-angle scattering. Close to the Sun, pitch-angle focusing still dominates, again resulting in a narrow strahl. Farther from the Sun, however, pitch-angle scattering dominates because focusing is effectively weakened by the increasing angle between the magnetic field direction and intensity gradient, a result of the spiral field. We determine the amount of scattering required to match Ulysses observations of strahl width in the fast solar wind, providing an important tool for inferring the large-scale properties and topologies of field lines in the interplanetary medium. Although the pitch-angle scattering term is independent of energy, time-of-flight effects in the spiral geometry result in an energy dependence of the strahl width that is in the observed sense although weaker in magnitude.

A comparative study of the effects of metal contamination on collembola in the field and in the laboratory

We examined the species diversity and abundance of Collembola at 32 sampling points along a gradient of metal contamination in a rough grassland site ( Wolverhampton, England), formerly used for the disposal of metal-rich smelting waste. Differences in the concentrations of Cd, Cu, Pb and Zn between the least and most contaminated part of the 35 metre transect were more than one order of magnitude. A gradient of Zn concentrations from 597 to 9080 mug g(-1) dry soil was found. A comparison between field concentrations of the four metals and previous studies on their relative toxicities to Collembola, suggested that Zn is likely to be responsible for any ecotoxicological effects on springtails at this site. Euedaphic ( soil dwelling) Collembola were extracted by placing soil cores into Tullgren funnels and epedaphic ( surface dwelling) species were sampled using pitfall traps. There was no obvious relationship between the total abundance, or a range of commonly used diversity indices, and Zn levels in soils. However, individual species showed considerable differences in abundance. Metal "tolerant'' (e.g., Ceratophysella denticulata) and metal "sensitive'' (e.g., Cryptopygus thermophilus) species could be identified. Epedaphic species appeared to be influenced less by metal contamination than euedaphic species. This difference is probably due to the higher mobility and lower contact with the soil pore water of epedaphic springtails in comparison to euedaphic Collembola. In an experiment exposing the standard test springtail, Folsomia candida, to soils from all 32 sampling points, adult survival and reproduction showed small but significant negative relationships with total Zn concentrations. Nevertheless, juveniles were still produced from eggs laid by females in the most contaminated soils with 9080 mug g(-1) Zn. Folsomia candida is much more sensitive to equivalent concentrations of Zn in the standard OECD soil. Thus, care should be taken in extrapolating the results of laboratory toxicity tests on metals in OECD soil to field soils, in which, the biological availability of contaminants is likely to be lower. Our studies have shown the importance of ecotoxicological effects at the species level. Although there may be no differences in overall abundance, sensitive species that are numerous in contaminated sites, and which may play important roles in decomposition("keystone species'') can be greatly reduced in numbers by pollution.

A comparative study of the effects of metal contamination on collembola in the field and in the laboratory

We examined the species diversity and abundance of Collembola at 32 sampling points along a gradient of metal contamination in a rough grassland site ( Wolverhampton, England), formerly used for the disposal of metal-rich smelting waste. Differences in the concentrations of Cd, Cu, Pb and Zn between the least and most contaminated part of the 35 metre transect were more than one order of magnitude. A gradient of Zn concentrations from 597 to 9080 mug g(-1) dry soil was found. A comparison between field concentrations of the four metals and previous studies on their relative toxicities to Collembola, suggested that Zn is likely to be responsible for any ecotoxicological effects on springtails at this site. Euedaphic ( soil dwelling) Collembola were extracted by placing soil cores into Tullgren funnels and epedaphic ( surface dwelling) species were sampled using pitfall traps. There was no obvious relationship between the total abundance, or a range of commonly used diversity indices, and Zn levels in soils. However, individual species showed considerable differences in abundance. Metal "tolerant'' (e.g., Ceratophysella denticulata) and metal "sensitive'' (e.g., Cryptopygus thermophilus) species could be identified. Epedaphic species appeared to be influenced less by metal contamination than euedaphic species. This difference is probably due to the higher mobility and lower contact with the soil pore water of epedaphic springtails in comparison to euedaphic Collembola. In an experiment exposing the standard test springtail, Folsomia candida, to soils from all 32 sampling points, adult survival and reproduction showed small but significant negative relationships with total Zn concentrations. Nevertheless, juveniles were still produced from eggs laid by females in the most contaminated soils with 9080 mug g(-1) Zn. Folsomia candida is much more sensitive to equivalent concentrations of Zn in the standard OECD soil. Thus, care should be taken in extrapolating the results of laboratory toxicity tests on metals in OECD soil to field soils, in which, the biological availability of contaminants is likely to be lower. Our studies have shown the importance of ecotoxicological effects at the species level. Although there may be no differences in overall abundance, sensitive species that are numerous in contaminated sites, and which may play important roles in decomposition("keystone species'') can be greatly reduced in numbers by pollution.

Variability in surface atmospheric electric field measurements

An electrical current of the order one picoamp per metre squared flows vertically in the Earth's atmosphere, between the ionosphere at approximately 50km altitude and the surface. This current is generated by global thunderstorm activity and is modulated by galactic cosmic rays and atmospheric aerosol. In fair weather conditions, this current cause a vertical atmospheric electric field, commonly measured as a potential gradient. For circumstances other than fair weather conditions, the potential gradient varies, from small steady enhancements in fog to large fluctuations in thunderstorms. The atmospheric potential gradient is continuously monitored at the Reading University Atmospheric Observatory. An account of the variability of the potential gradient on a variety of time scales will be presented.

Contrasting arbuscular mycorrhizal communities colonizing different host plants show a similar response to a soil phosphorus concentration gradient

High soil phosphorus (P) concentration is frequently shown to reduce root colonization by arbuscular mycorrhizal (AM) fungi, but the influence of P on the diversity of colonizing AM fungi is uncertain. We used terminal restriction fragment length polymorphism (T-RFLP) of 18S rDNA and cloning to assess diversity of AM fungi colonizing maize (Zea mays), soybean (Glycene max) and field violet (Viola arvensis) at three time points in one season along a P gradient of 10280mgl1 in the field. Percentage AM colonization changed between sampling time points but was not reduced by high soil P except in maize. There was no significant difference in AM diversity between sampling time points. Diversity was reduced at concentrations of P > 25mgl1, particularly in maize and soybean. Both cloning and T-RFLP indicated differences between AM communities in the different host species. Host species was more important than soil P in determining the AM community, except at the highest P concentration. Our results show that the impact of soil P on the diversity of AM fungi colonizing plants was broadly similar, despite the fact that different plants contained different communities. However, subtle differences in the response of the AM community in each host were evident.