Approximate maximum entropy joint feature inference consistent with arbitrary lower-order probability constraints: application to statistical classification

We propose a new learning method for discrete space statistical classifiers. Similar to Chow and Liu (1968) and Cheeseman (1983), we cast classification/inference within the more general framework of estimating the joint probability mass function (p.m.f.) for the (feature vector, class label) pair. Cheeseman's proposal to build the maximum entropy (ME) joint p.m.f. consistent with general lower-order probability constraints is in principle powerful, allowing general dependencies between features. However, enormous learning complexity has severely limited the use of this approach. Alternative models such as Bayesian networks (BNs) require explicit determination of conditional independencies. These may be difficult to assess given limited data. Here we propose an approximate ME method, which, like previous methods, incorporates general constraints while retaining quite tractable learning. The new method restricts joint p.m.f. support during learning to a small subset of the full feature space. Classification gains are realized over dependence trees, tree-augmented naive Bayes networks, BNs trained by the Kutato algorithm, and multilayer perceptrons. Extensions to more general inference problems are indicated. We also propose a novel exact inference method when there are several missing features.

Supercritical fluid extraction and accelerated solvent extraction of dioxins from high- and low-carbon fly ash

This study investigates the replacement of Soxhlet extraction by supercritical fluid extraction (SFE) or accelerated solvent extraction (ASE) for the removal of dioxins from municipal waste incinerator fly ash. SFE is very matrix dependent; higher percent recoveries versus Soxhlet extraction can be obtained for low-carbon-level fly ash, but only a few percent of dioxins can be extracted from high-carbon-level fly ash. The addition of large quantities of toluene in the extraction cell prior to extraction of high-carbon fly ash improves the recovery of the lowest chlorinated dioxins (approximately 90%), but a maximum of 20% of the octachlorodibenzo-p-dioxins can be extracted. Since large quantities of toluene are needed to improve the recoveries, ASE with toluene was tested. Recoveries similar to Soxhlet extraction can be obtained in 2 h at 80 degrees C. Increasing the temperature to 150 degrees C increases the extraction rate and yields recoveries of approximately 110-160% compared to 48-h Soxhlet extraction for all congeners for both low- and high-carbon fly ashes. These results question the choice of Soxhlet extraction as a reference method for dioxin determination.

Catalytic Upgrading of Fermentation-Derived Organic Acids.

The production of organic acids in low-cost, high-efficiency fermentation processes makes available a new route to chemical production from biomass. Because of their multiple functional groups and high reactivity, organic acids can undergo a variety of reactions that are effectively catalyzed by inorganic heterogeneous or homogeneous catalysts. Lactic acid and succinic acid, in particular, are approaching large-scale production via fermentation and show excellent promise as feedstocks for catalytic conversion routes such as hydrogenation, dehydration, or condensation. A number of catalytic conversion pathways of organic acids are potentially competitive with petroleum-based routes in the current economic environment, particularly when integrated into existing biomass/crop processing schemes. This article reviews some of the key reaction pathways available using fermentation-derived organic acids as feedstocks and presents recent results from the authors' lab on succinate hydrogenation to 1,4-butanediol and tetrahydrofuran. By a judicious choice of support properties and reaction conditions, it is possible to achieve yields of either of these two products in excess of 80%.

Light-regulated transcription of genes encoding peridinin chlorophyll a proteins and the major intrinsic light-harvesting complex proteins in the dinoflagellate amphidinium carterae hulburt (Dinophycae). Changes In cytosine methylation accompany photoadaptation

In the dinoflagellate Amphidinium carterae, photoadaptation involves changes in the transcription of genes encoding both of the major classes of light-harvesting proteins, the peridinin chlorophyll a proteins (PCPs) and the major a/c-containing intrinsic light-harvesting proteins (LHCs). PCP and LHC transcript levels were increased up to 86- and 6-fold higher, respectively, under low-light conditions relative to cells grown at high illumination. These increases in transcript abundance were accompanied by decreases in the extent of methylation of CpG and CpNpG motifs within or near PCP- and LHC-coding regions. Cytosine methylation levels in A. carterae are therefore nonstatic and may vary with environmental conditions in a manner suggestive of involvement in the regulation of gene expression. However, chemically induced undermethylation was insufficient in activating transcription, because treatment with two methylation inhibitors had no effect on PCP mRNA or protein levels. Regulation of gene activity through changes in DNA methylation has traditionally been assumed to be restricted to higher eukaryotes (deuterostomes and green plants); however, the atypically large genomes of dinoflagellates may have generated the requirement for systems of this type in a relatively "primitive" organism. Dinoflagellates may therefore provide a unique perspective on the evolution of eukaryotic DNA-methylation systems.