Regularisation of mixture density networks

Mixture Density Networks are a principled method to model conditional probability density functions which are non-Gaussian. This is achieved by modelling the conditional distribution for each pattern with a Gaussian Mixture Model for which the parameters are generated by a neural network. This thesis presents a novel method to introduce regularisation in this context for the special case where the mean and variance of the spherical Gaussian Kernels in the mixtures are fixed to predetermined values. Guidelines for how these parameters can be initialised are given, and it is shown how to apply the evidence framework to mixture density networks to achieve regularisation. This also provides an objective stopping criteria that can replace the `early stopping' methods that have previously been used. If the neural network used is an RBF network with fixed centres this opens up new opportunities for improved initialisation of the network weights, which are exploited to start training relatively close to the optimum. The new method is demonstrated on two data sets. The first is a simple synthetic data set while the second is a real life data set, namely satellite scatterometer data used to infer the wind speed and wind direction near the ocean surface. For both data sets the regularisation method performs well in comparison with earlier published results. Ideas on how the constraint on the kernels may be relaxed to allow fully adaptable kernels are presented.

Application of Mixture Models to Survival Data

Survival models are being widely applied to the engineering field to model time-to-event data once censored data is here a common issue. Using parametric models or not, for the case of heterogeneous data, they may not always represent a good fit. The present study relays on critical pumps survival data where traditional parametric regression might be improved in order to obtain better approaches. Considering censored data and using an empiric method to split the data into two subgroups to give the possibility to fit separated models to our censored data, we’ve mixture two distinct distributions according a mixture-models approach. We have concluded that it is a good method to fit data that does not fit to a usual parametric distribution and achieve reliable parameters. A constant cumulative hazard rate policy was used as well to check optimum inspection times using the obtained model from the mixture-model, which could be a plus when comparing with the actual maintenance policies to check whether changes should be introduced or not.

Removal of toxic metals in a multi metal system using sorbents for potential application to urban stormwater treatment

In the context of increasing demand for potable water and the depletion of water resources, stormwater is a logical alternative. However, stormwater contains pollutants, among which metals are of particular interest due to their toxicity and persistence in the environment. Hence, it is imperative to remove toxic metals in stormwater to the levels prescribed by drinking water guidelines for potable use. Consequently, various techniques have been proposed, among which sorption using low cost sorbents is economically viable and environmentally benign in comparison to other techniques. However, sorbents show affinity towards certain toxic metals, which results in poor removal of other toxic metals. It was hypothesised in this study that a mixture of sorbents that have different metal affinity patterns can be used for the efficient removal of a range of toxic metals commonly found in stormwater. The performance of six sorbents in the sorption of Al, Cr, Cu, Pb, Ni, Zn and Cd, which are the toxic metals commonly found in urban stormwater, was investigated to select suitable sorbents for creating the mixtures. For this purpose, a multi criteria analytical protocol was developed using the decision making methods: PROMETHEE (Preference Ranking Organisation METHod for Enrichment Evaluations) and GAIA (Graphical Analysis for Interactive Assistance). Zeolite and seaweed were selected for the creation of trial mixtures based on their metal affinity pattern and the performance on predetermined selection criteria. The metal sorption mechanisms employed by seaweed and zeolite were defined using kinetics, isotherm and thermodynamics parameters, which were determined using the batch sorption experiments. Additionally, the kinetics rate-limiting steps were identified using an innovative approach using GAIA and Spearman correlation techniques developed as part of the study, to overcome the limitation in conventional graphical methods in predicting the degree of contribution of each kinetics step in limiting the overall metal removal rate. The sorption kinetics of zeolite was found to be primarily limited by intraparticle diffusion followed by the sorption reaction steps, which were governed mainly by the hydrated ionic diameter of metals. The isotherm study indicated that the metal sorption mechanism of zeolite was primarily of a physical nature. The thermodynamics study confirmed that the energetically favourable nature of sorption increased in the order of Zn < Cu < Cd < Ni < Pb < Cr < Al, which is in agreement with metal sorption affinity of zeolite. Hence, sorption thermodynamics has an influence on the metal sorption affinity of zeolite. On the other hand, the primary kinetics rate-limiting step of seaweed was the sorption reaction process followed by intraparticle diffusion. The boundary layer diffusion was also found to limit the metal sorption kinetics at low concentration. According to the sorption isotherm study, Cd, Pb, Cr and Al were sorbed by seaweed via ion exchange, whilst sorption of Ni occurred via physisorption. Furthermore, ionic bonding is responsible for the sorption of Zn. The thermodynamics study confirmed that sorption by seaweed was energetically favourable in the order of Zn < Cu < Cd < Cr . Al < Pb < Ni. However, this did not agree with the affinity series derived for seaweed suggesting a limited influence of sorption thermodynamics on metal affinity for seaweed. The investigation of zeolite-seaweed mixtures indicated that mixing sorbents have an effect on the kinetics rates and the sorption affinity. Additionally, the theoretical relationships were derived to predict the boundary layer diffusion rate, intraparticle diffusion rate, the sorption reaction rate and the enthalpy of mixtures based on that of individual sorbents. In general, low coefficient of determination (R2) for the relationships between theoretical and experimental data indicated that the relationships were not statistically significant. This was attributed to the heterogeneity of the properties of sorbents. Nevertheless, in relative terms, the intraparticle diffusion rate, sorption reaction rate and enthalpy of sorption had higher R2 values than the boundary layer diffusion rate suggesting that there was some relationship between the former set of parameters of mixtures and that of sorbents. The mixture, which contained 80% of zeolite and 20% of seaweed, showed similar affinity for the sorption of Cu, Ni, Cd, Cr and Al, which was attributed to approximately similar sorption enthalpy of the metal ions. Therefore, it was concluded that the seaweed-zeolite mixture can be used to obtain the same affinity for various metals present in a multi metal system provided the metal ions have similar enthalpy during sorption by the mixture.

Mixture models of nucleotide sequence evolution that account for heterogeneity in the substitution process across sites and across lineages

Molecular phylogenetic studies of homologous sequences of nucleotides often assume that the underlying evolutionary process was globally stationary, reversible, and homogeneous (SRH), and that a model of evolution with one or more site-specific and time-reversible rate matrices (e.g., the GTR rate matrix) is enough to accurately model the evolution of data over the whole tree. However, an increasing body of data suggests that evolution under these conditions is an exception, rather than the norm. To address this issue, several non-SRH models of molecular evolution have been proposed, but they either ignore heterogeneity in the substitution process across sites (HAS) or assume it can be modeled accurately using the distribution. As an alternative to these models of evolution, we introduce a family of mixture models that approximate HAS without the assumption of an underlying predefined statistical distribution. This family of mixture models is combined with non-SRH models of evolution that account for heterogeneity in the substitution process across lineages (HAL). We also present two algorithms for searching model space and identifying an optimal model of evolution that is less likely to over- or underparameterize the data. The performance of the two new algorithms was evaluated using alignments of nucleotides with 10 000 sites simulated under complex non-SRH conditions on a 25-tipped tree. The algorithms were found to be very successful, identifying the correct HAL model with a 75% success rate (the average success rate for assigning rate matrices to the tree's 48 edges was 99.25%) and, for the correct HAL model, identifying the correct HAS model with a 98% success rate. Finally, parameter estimates obtained under the correct HAL-HAS model were found to be accurate and precise. The merits of our new algorithms were illustrated with an analysis of 42 337 second codon sites extracted from a concatenation of 106 alignments of orthologous genes encoded by the nuclear genomes of Saccharomyces cerevisiae, S. paradoxus, S. mikatae, S. kudriavzevii, S. castellii, S. kluyveri, S. bayanus, and Candida albicans. Our results show that second codon sites in the ancestral genome of these species contained 49.1% invariable sites, 39.6% variable sites belonging to one rate category (V1), and 11.3% variable sites belonging to a second rate category (V2). The ancestral nucleotide content was found to differ markedly across these three sets of sites, and the evolutionary processes operating at the variable sites were found to be non-SRH and best modeled by a combination of eight edge-specific rate matrices (four for V1 and four for V2). The number of substitutions per site at the variable sites also differed markedly, with sites belonging to V1 evolving slower than those belonging to V2 along the lineages separating the seven species of Saccharomyces. Finally, sites belonging to V1 appeared to have ceased evolving along the lineages separating S. cerevisiae, S. paradoxus, S. mikatae, S. kudriavzevii, and S. bayanus, implying that they might have become so selectively constrained that they could be considered invariable sites in these species.

Gibbs energies of formation of rare earth oxysulfides

The standard Gibbs energy change accompanying the conversion of rare earth oxides to oxysulfides by reaction of rare earth oxides with diatomic sulfur gas has been measured in the temperature range 870 to 1300 K using the solid state cell: Pt/Cu+Cu2S/R2O2S+R2O3‖(CaO)ZrO2‖Ni+NiO, Pt where R=La, Nd, Sm, Gd, Tb, and Dy. The partial pressure of diatomic sulfur over a mixture of rare earth oxide (R2O3) and oxysulfide (R2O2S) is fixed by the dissociation of Cu2S to Cu in a closed system. The buffer mixture of Cu+Cu2S is physically separated from the rare earth oxide and oxysulfide to avoid complications arising from interaction between them. The corresponding equilibrium oxygen partial pressure is measured with an oxide solid electrolyte cell. Gibbs energy change for the conversion of oxide to the corresponding oxysulfide increases monotonically with atomic number of the rare earth element. Second law enthalpy of formation also shows a similar trend. Based on this empirical trend Gibbs energies of formation of oxysulfides of Pr, Eu, Ho, and Er are estimated as a function of temperature.

Phase diagram and dielectric studies of binary organic materials

The phase equilibrium studies of organic system, involving resorcinol (R) and p-dimethylaminobenzaldehyde (DMAB), reveal the formation of a 1:1 molecular complex with two eutectics. The heat of mixing, entropy of fusion, roughness parameter, interfacial energy, and the excess thermodynamic functions were calculated based on enthalpy of fusion data determined via differential scanning calorimetric (DSC) method. X-ray powder diffraction studies confirm that the eutectics are not simple mechanical mixture of the components under investigation. The spectroscopic investigations (IR and NMR) suggest the occurrence of hydrogen bonding between the components forming the molecular complex. The dielectric measurements, carried out on hot-pressed addition compound (molecular complex), show higher dielectric constant at 320 K than that of individual components. The microstructural investigations of eutectic and addition compound indicate dendritic and faceted morphological features. (C) 2000 Elsevier Science B.V. All rights reserved.

A Thermodynamic Proposal for the Exposition of Critically and other Subtle Features in Self-Deflagrating Ammonium Perchlorate Systems

A first comprehensive investigation on the deflagration of ammonium perchlorate (AP) in the subcritical regime, below the low pressure deflagration limit (LPL, 2.03 MPa) christened as regime I$^{\prime}$, is discussed by using an elegant thermodynamic approach. In this regime, deflagration was effected by augmenting the initial temperature (T$_{0}$) of the AP strand and by adding fuels like aliphatic dicarboxylic acids or polymers like carboxy terminated polybutadiene (CTPB). From this thermodynamic model, considering the dependence of burning rate ($\dot{r}$) on pressure (P) and T$_{0}$, the true condensed (E$_{\text{s,c}}$) and gas phase (E$_{\text{s,g}}$) activation energies, just below and above the surface respectively, have been obtained and the data clearly distinguishes the deflagration mechanisms in regime I$^{\prime}$ and I (2.03-6.08 MPa). Substantial reduction in the E$_{\text{s,c}}$ of regime I$^{\prime}$, compared to that of regime I, is attributed to HClO$_{4}$ catalysed decomposition of AP. HClO$_{4}$ formation, which occurs only in regime I$^{\prime}$, promotes dent formation on the surface as revealed by the reflectance photomicrographs, in contrast to the smooth surface in regime I. The HClO$_{4}$ vapours, in regime I$^{\prime}$, also catalyse the gas phase reactions and thus bring down the E$_{\text{s,g}}$ too. The excess heat transferred on to the surface from the gas phase is used to melt AP and hence E$_{\text{s,c}}$, in regime I, corresponds to the melt AP decomposition. It is consistent with the similar variation observed for both the melt layer thickness and $\dot{r}$ as a function of P. Thermochemical calculations of the surface heat release support the thermodynamic model and reveal that the AP sublimation reduces the required critical exothermicity of 1108.8 kJ kg$^{-1}$ at the surface. It accounts for the AP not sustaining combustion in the subcritical regime I$^{\prime}$. Further support for the model comes from the temperature-time profiles of the combustion train of AP. The gas and condensed phase enthalpies, derived from the profile, give excellent agreement with those computed thermochemically. The $\sigma _{\text{p}}$ expressions derived from this model establish the mechanistic distinction of regime I$^{\prime}$ and I and thus lend support to the thermodynamic model. On comparing the deflagration of strand against powder AP, the proposed thermodynamic model correctly predicts that the total enthalpy of the condensed and gas phases remains unaltered. However, 16% of AP particles undergo buoyant lifting into the gas phase in the `free board region' (FBR) and this renders the demarcation of the true surface difficult. It is found that T$_{\text{s}}$ lies in the FBR and due to this, in regime I$^{\prime}$, the E$_{\text{s,c}}$ of powder AP matches with the E$_{\text{s,g}}$ of the pellet. The model was extended to AP/dicarboxylic acids and AP/CTPB mixture. The condensed ($\Delta $H$_{1}$) and gas phase ($\Delta $H$_{2}$) enthalpies were obtained from the temperature profile analyses which fit well with those computed thermochemically. The $\Delta $H$_{1}$ of the AP/succinic acid mixture was found just at the threshold of sustaining combustion. Indeed the lower homologue malonic acid, as predicted, does not sustain combustion. In vaporizable fuels like sebacic acid the E$_{\text{s,c}}$ in regime I$^{\prime}$, understandably, conforms to the AP decomposition. However, the E$_{\text{s,c}}$ in AP/CTPB system corresponds to the softening of the polymer which covers AP particles to promote extensive condensed phase reactions. The proposed thermodynamic model also satisfactorily explains certain unique features like intermittent, plateau and flameless combustion in AP/ polymeric fuel systems.

Standard Gibbs energy of formation of Pb2Ru2O6.5

Lead ruthenate is used as a bifunctional electrocatalyst for both oxygen evolution and reduction and as a conducting component in thick-film resistors. It also has potential applications in supercapacitors and solid oxide fuel cells. However, thermodynamic properties of the compound have not been reported in the literature. The standard Gibbs energy of formation has now been determined in the temperature range from 873 to 1123 K using a solid-state cell incorporating yttria-stabilized zirconia (YSZ) as the electrolyte, a mixture of PbO + Pb2Ru2O6.5 + Ru as the measuring electrode, and Ru + RuO2 as the reference. The design of the measuring electrode is based on a study of phase relations in the ternary system Pb–Ru–O at 1123 K. For the reaction,S0884291400095625_eqnU1 the standard enthalpy of formation and standard entropy at 298.15 K are estimated from the high-temperature measurements. An oxygen potential diagram for the system Pb–Ru–O is composed based on data obtained in this study and auxiliary information from the literature

Measurement of the Gibbs energy of formation of URh3 using an oxide solid electrolyte

The Gibbs' energy offormation of the intermetallic compound URh3has been measured in the temperature range 980 to 1320 K using an oxide solid state cell incorporating yttria-doped thoria as the solid electrolyte and a mixture of manganese and manganese oxide as the reference electrode. The cell can be represented as Pt, Mn + MnO I (Y203)Th02 I Rh + URh3 + U02 + x' Rh, Pt The reversible emf of the cell was a linear function of temperature E = 15.60 +0.0237 T (±0.8) mY. Using auxiliary thermodynamic data for MnO and U02+ x the Gibbs' energy of formation of URh3 from component metals has been computed. The results can be expressed by the equation L'.G?< URh3 > = -316240 + 13.22 T (± 3000) J mol-1. The "third-law" enthalpy of formation of URh3at 298 K is -293.2 (± 4) kJ mol-1, significantly more negative than the value of -181.5 kJ mol-1 calculated using Miedema's model.

The standard enthalpy and entropy of formation of Rh2O3 - A third-law optimization

The standard Gibbs energy of formation of Rh203 at high temperature has been determined recently with high precision. The new data are significantly different from those given in thermodynamic compilations.Accurate values for enthalpy and entropy of formation at 298.15 K could not be evaluated from the new data,because reliable values for heat capacity of Rh2O3 were not available. In this article, a new measurement of the high temperature heat capacity of Rh2O3 using differential scanning calorimetry (DSC) is presented.The new values for heat capacity also differ significantly from those given in compilations. The information on heat capacity is coupled with standard Gibbs energy of formation to evaluate values for standard enthalpy and entropy of formation at 289.15 K using a multivariate analysis. The results suggest a major revision in thermodynamic data for Rh2O3. For example, it is recommended that the standard entropy of Rh203 at 298.15 K be changed from 106.27 J mol-' K-'given in the compilations of Barin and Knacke et al. to 75.69 J mol-' K". The recommended revision in the standard enthalpy of formation is from -355.64 kJ mol-'to -405.53 kJ mol".

Phase Equilibria in the System Al2O3-CaO-CoO and Gibbs Energy of Formation of Ca3CoAl4O10

An isothermal section of the system Al2O3-CaO-CoO at 1500 K has been established by equilibrating 22 samples of different compositions at high temperature and phase identification by optical and scanning electron microscopy, X-ray diffraction, and energy dispersive spectroscopy after quenching to room temperature. Only one quaternary oxide, Ca3CoAl4O10, was identified inside the ternary triangle. Based on the phase relations, a solid-state electrochemical cell was designed to measure the Gibbs energy of formation of Ca3CoAl4O10 in the temperature range from 1150 to 1500 K. Calcia-stabilized zirconia was used as the solid electrolyte and a mixture of Co + CoO as the reference electrode. The cell can be represented as: ( - )\textPt,\textCaAl 2 \textO 4 + \textCa 1 2 \textAl 1 4 \textO 3 3 + \textCa 3 \textCoAl 4 \textO 10 + \textCo//(CaO)ZrO 2 \text// \textCoO + \textCo,\text Pt ( + ). (−)PtCaAl2O4+Ca12Al14O33+Ca3CoAl4O10+Co//(CaO)ZrO2//CoO+Co Pt (+) From the emf of the cell, the standard Gibbs energy change for the Ca3CoAl4O10 formation reaction, CoO + 3/5CaAl2O4 + 1/5Ca12Al14O33 → Ca3CoAl4O10, is obtained as a function of temperature: \Updelta Gr\texto Unknown control sequence '\Updelta'/J mol−1 (±50) = −2673 + 0.289 (T/K). The standard Gibbs energy of formation of Ca3CoAl4O10 from its component binary oxides, Al2O3, CaO, and CoO is derived as a function of temperature. The standard entropy and enthalpy of formation of Ca3CoAl4O10 at 298.15 K are evaluated. Chemical potential diagrams for the system Al2O3-CaO-CoO at 1500 K are presented based on the results of this study and auxiliary information from the literature.

Study of the effect of alkali mixture on V-O bond length in Oxyfluoro Vanadate glasses using Raman spectroscopy

Raman spectroscopic study on Oxyfluoro Vanadate glasses containing various proportions of lithium fluoride and rubidium fluoride was carried out to see an effect of mixture of alkali on vanadium-oxygen (V-O) bond length. Glasses with a general formula 40V(2)O(5) - 30BaF(2) - (30 - x) LiF - xRbF (x = 0-30) were prepared. Room temperature Raman spectra of these glass samples were recorded in back scattering geometry. The data presented is in ``reduced Raman intensity'' form with maximum peak scaled to 100. We have used v = Aexp(BR), where A and B are fitting parameters, to correlate the bond length R with Raman scattering frequency v. We observed that variation in bond length and its distribution about a most probable value can be correlated to the alkali environment present in these glasses. We also observed that all rubidium environment around the network forming unit is more homogenous than all lithium environment.

Gibbs energy of formation of Ca7V4O17

Thermodynamic properties of Ca7V4O17 are measured for the first time using a solid-state electrochemical cell incorporating single crystal of CaF2 as the electrolyte over the temperature range from (900 to 1175) K. An equimolar mixture of CaO and CaF2 is used as the reference electrode and a mixture of Ca3V2O8, Ca7V4O17 and CaF2 as the measuring electrode. Both the electrodes are placed under flowing oxygen gas at ambient pressure. The standard Gibbs energy change for the reaction: 2Ca(3)V(2)O(8) + CaO -> Ca7V4O17; which is related to the chemical potential of CaO in the two-phase region (Ca3V2O8 + Ca7V4O17) of the pseudo-binary system CaO + V2O5, is obtained from the electromotive force of the cell as: Delta(r)G(o) +/- 127/(J . mol(-1)) = Delta mu(CaO) = -11453 + 8.273(T/K). The derived standard enthalpy of formation of Ca7V4O17 from elements in their normal standard states is ( 8208.97 +/- 8) kJ . mol (1) and its standard entropy is (560.05 +/- 7.5) J . K (1) . mol (1), both at T = 298.15 K. The results indicate that Ca7V4O17 decomposes into Ca3V2O8 and CaO at T = (1384 +/- 3) K.

Gibbs free energy of formation of rhodium sulfides

Using a solid-state electrochemical technique, thermodynamic properties of three sulfide phases (RhS0.882, Rh3S4, Rh2S3) in the binary system (Rh + S) are measured as a function of temperature over the range from (925 to 1275) K. Single crystal CaF2 is used as the electrolyte. The auxiliary electrode consisting of (CaS + CaF2) is designed in such a way that the sulfur chemical potential converts into an equivalent fluorine potential at each electrode. The sulfur potentials at the measuring electrodes are established by the mixtures of (Rh + RhS0.882), (RhS0.882 + Rh3S4) and (Rh3S4 + Rh2S3) respectively. A gas mixture (H-2 + H2S + Ar) of known composition fixes the sulfur potential at the reference electrode. A novel cell design with physical separation of rhodium sulfides in the measuring electrode from CaS in the auxiliary electrode is used to prevent interaction between the two sulfide phases. They equilibrate only via the gas phase in a hermetically sealed reference enclosure. Standard Gibbs energy changes for the following reactions are calculated from the electromotive force of three cells: 2.2667Rh (s) + S-2 (g) -> 2.2667RhS(0.882) (s), Delta(r)G degrees +/- 2330/(J . mol(-1)) = -288690 + 146.18 (T/K), 4.44RhS(0.882) (s) + S-2 (g) -> 1.48Rh(3)S(4) (s), Delta(r)G degrees +/- 2245/(J . mol(-1)) = -245596 + 164.31 (T/K), 4Rh(3)S(4) (s) + S-2 (g) -> 6Rh(2)S(3) (s), Delta(r)G degrees +/- 2490/(J . mol(-1)) = -230957 + 160: 03 (T/K). Standard entropy and enthalpy of formation of rhodium sulfides from elements in their normal standard states at T = 298.15 K are evaluated. (C) 2013 Elsevier Ltd. All rights reserved.

Thermodynamic Properties of LaCrO4, La2CrO6, and La2Cr3O12, and Subsolidus Phase Relations in the System Lanthanum-Chromium-Oxygen

In the system La-Cr-O, there are three ternary oxides (LaCrO4, La2Cr3O12, and La2CrO6) that contain Cr in higher valence states (V or VI). On heating, LaCrO4 decomposes to LaCrO3, La2Cr3O12 to a mixture of LaCrO4 and Cr2O3, and La2CrO6 to LaCrO3 and La2O3 with loss of oxygen. The oxygen potentials corresponding to these decomposition reactions are determined as a function of temperature using solid-state cells incorporating yttria-stabilized zirconia as the electrolyte. Measurements are made from 840K to the decomposition temperature of the ternary oxides in pure oxygen. The standard Gibbs energies of formation of the three ternary oxides are derived from the reversible electromotive force (EMF) of the three cells. The standard enthalpy of formation and standard entropy of the three ternary oxides at 298.15K are estimated. Subsolidus phase relations in the system La-Cr-O are computed from thermodynamic data and displayed as isothermal sections at several temperature intervals. The decomposition temperatures in air are 880 (+/- 3)K for La2Cr3O12, 936 (+/- 3)K for LaCrO4, and 1056 (+/- 4)K for La2CrO6.