We study discriminative clustering for market segmentation tasks. The underlying problem setting resembles discriminative clustering, however, existing approaches focus on the prediction of univariate cluster labels. By contrast, market segments encode complex (future) behavior of the individuals which cannot be represented by a single variable. In this paper, we generalize discriminative clustering to structured and complex output variables that can be represented as graphical models. We devise two novel methods to jointly learn the classifier and the clustering using alternating optimization and collapsed inference, respectively. The two approaches jointly learn a discriminative segmentation of the input space and a generative output prediction model for each segment. We evaluate our methods on segmenting user navigation sequences from Yahoo! News. The proposed collapsed algorithm is observed to outperform baseline approaches such as mixture of experts. We showcase exemplary projections of the resulting segments to display the interpretability of the solutions.

We study the problem of identifying botnets and the IP addresses which they comprise, based on the observation of a fraction of the global email spam traffic. Observed mailing campaigns constitute evidence for joint botnet membership, they are represented by cliques in the graph of all messages. No evidence against an association of nodes is ever available. We reduce the problem of identifying botnets to a problem of finding a minimal clustering of the graph of messages. We directly model the distribution of clusterings given the input graph; this avoids potential errors caused by distributional assumptions of a generative model. We report on a case study in which we evaluate the model by its ability to predict the spam campaign that a given IP address is going to participate in.

We study a setting in which Poisson processes generate sequences of decision-making events. The optimization goal is allowed to depend on the rate of decision outcomes; the rate may depend on a potentially long backlog of events and decisions. We model the problem as a Poisson process with a throttling policy that enforces a data-dependent rate limit and reduce the learning problem to a convex optimization problem that can be solved efficiently. This problem setting matches applications in which damage caused by an attacker grows as a function of the rate of unsuppressed hostile events. We report on experiments on abuse detection for an email service.

Abstract: Gold nanoparticles supported on TiO2 were probed by adsorption of methyl mercaptane (MM), and the process was quantified gravimetrically. This method allowed discrimination between weakly adsorbed (physisorbed) and strongly bound (chemisorbed) methyl mercaptane. Strong adsorption of MM occured on exposed Au faces, while low-temperature pre-treatment (30 degrees C) completely suppressed adsorption of MM on the TiO2 support. The thus obtained high selectivity of MM adsorption on Au enabled characterization of the gold surface area and the resulting values are comparable with other noble metal systems of similar average particle size. The estimated adsorption stoichiometry indicates that the entire Au surface is probed.

Well-characterized Au nanoparticles were deposited on commercial TiO(2) (P25, Degussa) and analyzed by means of STEM and thermogravimetry coupled with mass spectrometry (TG-MS). The adsorption was studied on Au/TiO(2) samples with Au loadings in the range of 1.1-9.9wt.% by injecting pulses of CH(3)SH (methyl mercaptane, MM) until no further mass increase could be observed. A prerequisite for determination of the surface area of the deposited gold nanoparticles is the proper discrimination of species adsorbing on the Au nanoparticles and the titania support. The adsorption of methyl mercaptane on the titania support strongly depended on the pretreatment temperature (30-400 degrees C), whereas the adsorption on Au nanoparticles was virtually unaffected by this parameter. A very mild thermal pretreatment was identified as a requirement for avoiding the adsorption of the MM on the titania support. CH(3)SH adsorbed on the support desorbed at lower temperatures (maximal rate of desorption was centered at ca. 150 degrees C) compared to species desorbing from Au nanoparticles (maximum at ca. 200-220 degrees C). Moreover, CH(3)SH adsorbed on Au nanoparticles desorbed in the form of dimethyl sulfide (CH(3))(2)S. Part of MM adsorbed on the gold surface was not desorbed even at high temperatures (above 500 degrees C) and stayed on the surface in the form of relatively stable C(x)H(y)S(z) fragments. This residue could be removed by oxygen pulses resulting in the formation of CO(2), SO(2), and H(2)O. The good discrimination of MM chemisorption on Au nanoparticles and on titania renders the determination of the Au surface area viable. Potential and limitations of the CH(3)SH chemisorption for the surface area determination of Au nanoparticles are discussed.

We discuss the problem of clustering elements according to the sources that have generated them. For elements that are characterized by independent binary attributes, a closed-form Bayesian solution exists. We derive a solution for the case of dependent attributes that is based on a transformation of the instances into a space of independent feature functions. We derive an optimization problem that produces a mapping into a space of independent binary feature vectors; the features can reflect arbitrary dependencies in the input space. This problem setting is motivated by the application of spam filtering for email service providers. Spam traps deliver a real-time stream of messages known to be spam. If elements of the same campaign can be recognized reliably, entire spam and phishing campaigns can be contained. We present a case study that evaluates Bayesian clustering for this application.

Selective blocking (poisoning) of catalytic active sites of gold catalysts, CeO2 and TiO2 supported Au colloids (2.1 and 6.9 nm), was investigated in the aerobic oxidation of benzyl alcohol as well as in the hydrogenation of ketopantolactone by adding various amounts of two chemically distinct thiols with different functionalities (n-octadecanethiol, ODT, and mercaptoacetic acid, MAA). There were clear trends in their poisoning behaviors, independent of the type of support and An particle size. ODT poisoned the catalyst in the oxidation reaction much stronger than MAA. In contrast, MAA was a stronger poisoning agent in the hydrogenation reaction. The characteristics Of Supported Au particles and the nature of thiol adsorption were investigated by Attenuated Total Reflection Infrared (ATR-IR) spectroscopy, TEM, thermogravimetric desorption and DFT calculations. Based on the adsorption characteristics of two thiols derived from the combined experimental and theoretical Study, it is Suggested that C=O hydrogenation of ketopantolactone occurs preferentially at low coordination Sites Such as corners and edges, which are more dominant in Small Clusters, while oxidation seems to be favored on extended active sites as prevailing on larger Au clusters. (C) 2009 Elsevier B.V. All rights reserved.

A parallel reactor cell for catalytic in situ X-ray absorption spectroscopy (XAS) measurements was developed. The cell facilitates the Simultaneous catalytic and structural investigation of six catalysts under different feed gas conditions. A two-dimensional X-ray sensor was used for spectra collection. Gas compositions were measured by on-line mass spectrometry. The potential and limitations of the high throughput XAS cell are discussed. The heterogeneously catalyzed partial oxidation of methane (CPO) was chosen as a test reaction. Alumina-supported Rh and Au/Rh catalysts with different metal loadings (0.5-2.5 wt%) were applied and prepared via different preparation routes using flame spray pyrolysis (fsp) and colloid adsorption (col). For comparison the same catalysts were also investigated in a fixed-bed capillary reactor heated by a gas blower (also for XAS measurements) and in an eight-fold parallel gas phase reactor using similar reaction conditions (6% CH4-3% O-2-He, 250-500 degrees C). Similar catalytic results were obtained in all three reactor types, confirming the suitability of the parallel reactor XAS cell for catalytic measurements. The catalysts showed different activity, selectivity and reducibility depending on metal loading, preparation route and Au/Rh ratio. The analysis of the catalytic data, STEM images and the in situ XANES experiments of the various catalysts indicated the following characteristics for the CPO reaction: sufficient heat production by combustion of methane, total conversion of oxygen, and reduction of a certain fraction of the catalyst. The overall catalytic behavior was in line with a two-zone model of the catalyst bed where catalytic combustion dominates in the front zone and reforming reactions become favored in the second part of the catalyst bed. (C) 2008 Elsevier B.V. All rights reserved.

The effect of the size of gold particles deposited on CeO 2 and TiO 2 supports on their catalytic behavior in the aerobic oxidation of benzyl alcohol in different solvents (mesitylene, toluene, and supercritical carbon dioxide) has been investigated. The size of supported gold particles deposited via a colloidal route was in the range 1.3–11.3 nm, as determined by means of EXAFS and HAADF-STEM measurements. The catalytic performance of the supported gold catalysts in the different solvents revealed a significant effect of the gold particle size. Optimal activity was observed for catalysts with medium particle size (ca. 6.9 nm) whereas smaller and bigger particles showed inferior activity. Identical trends for the activity–particle size relationship were found using Au/CeO 2 and Au/TiO 2 for the reaction at atmospheric pressure in conventional solvents (mesitylene, toluene) as well as under supercritical conditions (scCO 2 ). Selectivity to benzaldehyde was only weakly affected by the gold particle size and mainly depended on reaction conditions. In supercritical CO 2 (scCO 2 ) selectivity was higher than in the conventional solvents under atmospheric pressure. All catalysts tested with particle sizes ranging from 1.3 to 11.3 nm showed excellent selectivity of 99% or higher under supercritical conditions.

Aerobic oxidation of 1-phenylethanol was investigated over Au deposited on flame-derived Mg–Al and Cu–Mg–Al mixed oxides with different metal ratios. A maximum in acetophenone (1-phenyl-ethanone) yield was observed for catalysts based on both Cu–Mg–Al and Mg–Al mixed oxides depending on their composition. Special attention was given to the elucidation of the role of surface basicity and the influence of the preparation route on the particle size of Au. Adsorption of CO2 from the liquid phase combined with in situ ATR-IR and modulation excitation spectroscopy (MES) was applied to investigate differences in the surface properties of the mixed oxides as a function of the composition. Monodentate and bidentate carbonates were identified, the former being dominant on supports with high Cu contents. In order to obtain a rough quantification of the surface basicity, the retroaldolisation of 4-hydroxy-4-methyl-2-pentanone (diacetone alcohol, DAA) was chosen as a probe reaction indicating that a ratio Mg/Al=3 results in optimal surface basicity. Moreover, the addition of Cu only lead to a partial loss in retroaldolisation activity, indicating that also the copper sites form basic centers on the surface, however, slightly weaker ones than the corresponding Mg sites. The preparation routes applied (adsorption of colloid, deposition precipitation, and impregnation) lead to different gold particle sizes characterized by mean diameters of ≈2, ≈9 and ≈30nm, respectively. Catalytic tests using Au/Cu1Mg2Al1Ox catalysts with different mean gold particle size hint towards a particle size dependence of the aerobic oxidation of 1-phenylethanol, showing higher activity for the catalyst containing gold particles of ca. 9nm compared to those with 2 and 30nm particles, respectively.

We address the problem of learning decision functions from training data in which some attribute values are unobserved. This problem can arise, for instance, when training data is aggregated from multiple sources, and some sources record only a subset of attributes. We derive a generic joint optimization problem in which the distribution governing the missing values is a free parameter. We show that the optimal solution concentrates the density mass on finitely many imputations, and provide a corresponding algorithm for learning from incomplete data. We report on empirical results on benchmark data, and on the email spam application that motivates our work.

A new spectroscopic cell suitable for the analysis of heterogeneous catalysts by fluorescence EXAFS (extended X-ray absorption fine structure), transmission EXAFS and X-ray diffraction during in situ treatments and during catalysis is described. Both gas-phase and liquid-phase reactions can be investigated combined with on-line product analysis performed either by mass spectrometry or infrared spectroscopy. The set-up allows measurements from liquid-nitrogen temperature to 973 K. The catalysts are loaded preferentially as powders, but also as self-supporting wafers. The reaction cell was tested in various case studies demonstrating its flexibility and its wide applicability from model studies at liquid-nitrogen temperature to operando studies under realistic reaction conditions. Examples include structural studies during (i) the reduction of alumina-supported noble metal particles prepared by flame-spray pyrolysis and analysis of alloying in bimetallic noble metal particles (0.1%Pt-0.1%Pd/Al(2)O(3), 0.1%Pt-0.1%Ru/Al(2)O(3), 0.1%Pt-0.1%Rh/Al(2)O(3), 0.1%Au-0.1%Pd/Al(2)O(3)), (ii) reactivation of aged 0.8%Pt-16%BaO-CeO(2) NO(x) storage-reduction catalysts including the NO(x) storage/reduction cycle, and (iii) alcohol oxidation over gold catalysts (0.6%Au-20%CuO-CeO(2)).

We address the problem of detecting batches of emails that have been created according to the same template. This problem is motivated by the desire to filter spam more effectively by exploiting collective information about entire batches of jointly generated messages. The application matches the problem setting of supervised clustering, because examples of correct clusterings can be collected. Known decoding procedures for supervised clustering are cubic in the number of instances. When decisions cannot be reconsidered once they have been made --- owing to the streaming nature of the data --- then the decoding problem can be solved in linear time. We devise a sequential decoding procedure and derive the corresponding optimization problem of supervised clustering. We study the impact of collective attributes of email batches on the effectiveness of recognizing spam emails.

Gold nanoparticles were deposited on mixed oxides containing Cu, Mg, and Al in different ratios. The mixed-oxide supports were prepared by flame spray pyrolysis (FSP), resulting in agglomerated primary nanoparticles in the 10-15 nm range, onto which 6- to 9-nm gold particles were deposited by means of deposition-precipitation. The mixed-oxide-supported Au catalysts with noble metal loading of 0.6 +/- 0.17 wt% were investigated concerning their structural properties and tested in the aerobic liquid-phase oxidation of I-phenylethanol to phenyl-methyl ketone affording TOFs up to 1300 h(-1). The catalytic tests showed that the activity of these catalysts depends strongly on the composition of the support, with Cu and Mg being crucial components. Strongly enhanced catalytic activity was observed for gold supported on a ternary mixed oxide containing Cu, Mg, and Al at the molar ratio of 5:1:2. Extension of the catalytic tests to various structurally different alcohols indicated that the ternary mixed-oxide-supported gold catalyst has excellent catalytic properties in the aerobic oxidation of a broad range of structurally different alcohols, affording selectivities >= 98%. XANES revealed both reduced and oxidized Au species on the ternary mixed-oxide supports before and after the reaction. CO2 adsorption from the liquid phase combined with in situ ATR-IR and modulation excitation spectroscopy was applied to investigate differences in the basic surface properties of the mixed oxides. Monodentate and bidentate carbonates were identified, the former being dominant on ternary Cu-containing supports. (c) 2007 Elsevier Inc. All rights reserved.

Chemie Ingenieur TechnikVolume 79, Issue 9 p. 1305-1305 PosterFree Access Operando XAS- und IR-Spektroskopie: Au-Katalysatoren während der Alkoholoxidation P. Haider, P. Haider Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this authorJ.-D. Grunwaldt PD Dr., J.-D. Grunwaldt PD Dr. Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this authorA. Baiker Prof., A. Baiker Prof. baiker@chem.ethz.ch Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this author P. Haider, P. Haider Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this authorJ.-D. Grunwaldt PD Dr., J.-D. Grunwaldt PD Dr. Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this authorA. Baiker Prof., A. Baiker Prof. baiker@chem.ethz.ch Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Hönggerberg, HCI, CH-8093 ZürichSearch for more papers by this author First published: 18 September 2007 https://doi.org/10.1002/cite.200750152AboutPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat No abstract is available for this article. Volume79, Issue9Special Issue: ProcessNet Jahrestagung 2007September, 2007Pages 1305-1305 ReferencesRelatedInformation

Liquid-phase in situ XANES experiments were used to elucidate the oxidation state of differently supported Au catalysts during the selective oxidation of 1-phenylethanol with molecular oxygen. The catalysts investigated included (i) Au(0.6 wt%)/Cu3Mg3Al2Ox, (ii) Au(0.6 wt%)/Mg3AlOx, (iii) Au(2.9 wt%)/CeO2, and (iv) a copper-doped analog, Au(0.6 wt%)/CuO–CeO2 (20 wt% CuO). The XANES and EXAFS data were collected mainly in the fluorescence mode. Simultaneous determination of the structure (XANES) and the activity of the catalysts in the continuous-flow microreactor was achieved by following the oxidation state using XANES and the characteristic carbonyl vibration of the desired reaction product acetophenone using an IR spectrometer equipped with a transmission cell. Au/Cu3Mg3Al2Ox showed a small increase of conversion with time on stream accompanied by a slight reduction of the gold component. With Au/Mg3AlOx and Au/CuO–CeO2, a stronger reduction accompanied by increased catalytic activity was observed. The results of these mixed oxide-supported Au catalysts demonstrate that metallic gold is the main active species in catalytic aerobic alcohol oxidation. In contrast, a decrease in the product formation rate was observed for Au/CeO2 with ongoing reduction. Results obtained from factor analysis indicate that the decrease in activity of the Au/CeO2 with time on stream does not correlate with the much faster reduction of the catalyst and thus cannot be attributed to the increasing fraction of Au0 species. This finding corroborates the importance of recording XANES spectra for structural identification simultaneously with online determination of the catalytic performance.

The problem of finding duplicates in data is ubiquitous in data mining. We cast the problem of finding duplicates in sequential data into a poly-cut problem on a fully connected graph. The edge weights can be identified with parameterized pairwise similarities between objects that are optimized by structural support vector machines on labeled training sets. Our approach adapts the similarity measure to the data and is independent of the number of clusters. We present three large margin approximations of learning the pairwise similarities: an integrated QPformulation, a sequential multi-class approach and a pairwise classifier. We report on experimental results.