This paper presents empirical methods for enhancing the accuracy of inductive learning systems. It

addresses the problems of: learning propositional production rules in multi-class classification tasks in

noisy domains, maintaining continuous learning when confronted with new situations after the learning

phase is completed, and classifying an object when no rule is satisfied for it.

It is shown that interleaving the learning and performance-evaluation processes, allows accurate

classifications to be made on real-world data sets. The paper presents the system ARIS which

implements this approach, and it is shown that the resulting classifications are often more accurate than

those made by the non-refined knowledge bases.

The core design decision that lies behind ARIS is that it employs an ordering of the rules according to

their weight. A rule’s weight is learned by using Bayes’ theorem to calculate weights for the rule’s

conditions and to combine them. This model focuses the analysis of the knowledge base and assists the

refinement process significantly.

The system is non-interactive, it relies on heuristics to focus the refinement on those experiments that

appear to be most consistent with the refinement data set. The design framework of ARIS consists of a

tabular model for expressing rule weights, and the relationship between refinement cases and the rules

satisfied for each case to focus the refinement process. The system has been used to refine knowledge

bases created by ARIS itself, as well as to refine knowledge bases created by the RIPPER and C4.5

systems [6,25] in ten selected domains. Two main advantages have been observed. First, the ability to

gradually improve the knowledge base as the refinement proceeds. Second, is the ability to learn strong

rules utilising condition’s weight using minimum covering algorithm. Thus, by taking account of this

information we improve the applicability and quality of refinement

The effect of Aldrich humic acid (HA) on the mobility of {sup 137}Cs, {sup 85}Sr, {sup 152}Eu and {sup 239}Pu radionuclides was studied in Ca-montmorillonite suspensions. Verified 2-sites-2-species (2s2s) models correspond to an intensive interaction of all elements with humificated surface, what is in a remarkable contrast with the weak complexation of cesium and even strontium in solutions - the neutral ligand interaction constants {beta} (l/mol) are log {beta} < -9.9 and 7.56 {+-} 0.21 for Cs and Sr, respectively. The result for europium complexation in solution, log {beta} = 12.49 {+-} 0.18 is in a good agreement with literature data. For plutonium(IV) not only high proton competitive constant in solution was obtained, log {beta} = (-0.67 {+-} 0.32)+3pH, but also a strong chemisorption, which at high concentrations of humic acid (above 0.05 g/l) indicates the formation of bridge humate complexes of plutonium on the humificated surface. Logarithms of heterogeneous interaction constants ({beta}{sub 10}, l/g) of the elements with surface humic acid are 4.47 {+-} 0.23, 4.39 {+-} 0.08, and 6.40 {+-} 0.33 for Cs, Sr, and Eu(III), respectively, and the logarithm of the proton competitive constant ({beta}{sub 24}, l/g) for Pu(IV) -3.80 {+-}0.72. Distribution coefficients of humic acid and metal humates between 0.01 g HA/l solution and montmorillonite were derived as log K{sub d}(AH) = -1.04 {+-} 0.11, log K{sub d}(EuA) = 1.56 {+-} 0.11 and log K{sub d}(PuA) = 2.25 {+-} 0.04, while the values for Cs and Sr were obtained with very high uncertainty. Speciation of the elements on montmorillonite surface is illustrated as a function of equilibrium concentration of humic acid in solution and of pH.

Abstract

The high-performance size-exclusion chromatography (HPSEC) and radiochromatography (HPSERC) was used for the identification of radiocesium and radiostrontium interaction with humic acid. It was found that the behavior of humic acid on size-exclusion chromatography is sensitive to the salt concentration and pH of the mobile phase. At lower ionic strength and in acidic region of pH, the Aldrich humic acid exhibited three main fraction within the ranges > 760 kDa, 25-100 kDa and < 5 kDa. Radiocesium was found in the low-molecular fractions (< 1 kDa) of humic acids but radiostrontium interacts preferably with the fractions of humic acid of molecular weight within the range 2-5 kDa. (author) 27 refs.

: This paper describes 29Si, 27Al, 207Pb and 113Cd NMR studies of Pb- and Cd-exchanged LTA zeolites. The 113Cd NMR spectra identify for the first time two different coordination states for Cd2C in hydrated CdLTA, consistent with x-ray single crystal diffraction data for this zeolite. The effects of Cd2C exchange on the 29Si NMR spectra of LTA are also consistent with the structural data. Cd2C exchange perturbs the 27Al NMR spectrum of the zeolite, as does Pb2C exchange. Changes in the 29Si NMR spectra on Pb2C exchange are explained in terms of structural distortion of the framework. 207Pb NMR spectra of Pb2C exchanged LTA show a single broad signal indicating that Pb2C cations are rigidly located in the zeolite at relatively long Pb—O distances in the hydrated zeolites.

Iron substituted ETS-10 samples with different Fe/Ti ratios were synthesised hydrothermally. The occurrence of a doublet with the isomer shift of octahedral Fe(III) in the Mossbauer spectra.

Lead compounds still have numerous applications in our society, in spite of their known toxicity *. For example they are used in batteries and petroleum additives, paints and housing materials, catalysts and ceramics, glasses and semi-conductors. Elemental analysis will readily show the presence of lead but precise knowledge of the form it takes should be accessible with solid-state 2G7pb NMR, because NMR can probe the local environment of the lead atom(s) 2 ' 8 . The enormous chemical shift range, >15,000 ppm, already established for lead 9 suggests that small changes in its environment will lead to easily detected changes in chemical shift provided that the usual line narrowing techniques developed for solids, such as magic angle spinning (MAS), are successful. On the other hand, distortions from regular geometry around the lead atom will induce large changes in chemical shift anisotropy, and make detection more difficult as has been found for ^H g *0. The aim of this work has been to investigate parameters and variables required for the successful detection of this nucleus which has very favourable properties ( spin 1/2, 21% natural abundance) in the solid state. We have recently shown that 207pj, NM R spectra of solid lead(II) nitrate are significantly and uniformily dependent on temperature ?. Here we present results for some simple lead compounds as well as for model sytems related to soils such as leadexchanged zeolites H