Optimization of Robust Loss Functions for Weakly-Labeled Image Taxonomies

The recently proposed ImageNet dataset consists of several million images, each annotated with a single object category. These annotations may be imperfect, in the sense that many images contain multiple objects belonging to the label vocabulary. In other words, we have a multi-label problem but the annotations include only a single label (which is not necessarily the most prominent). Such a setting motivates the use of a robust evaluation measure, which allows for a limited number of labels to be predicted and, so long as one of the predicted labels is correct, the overall prediction should be considered correct. This is indeed the type of evaluation measure used to assess algorithm performance in a recent competition on ImageNet data. Optimizing such types of performance measures presents several hurdles even with existing structured output learning methods. Indeed, many of the current state-of-the-art methods optimize the prediction of only a single output label, ignoring this ‘structure’ altogether. In this paper, we show how to directly optimize continuous surrogates of such performance measures using structured output learning techniques with latent variables. We use the output of existing binary classifiers as input features in a new learning stage which optimizes the structured loss corresponding to the robust performance measure. We present empirical evidence that this allows us to ‘boost’ the performance of binary classification on a variety of weakly-supervised labeling problems defined on image taxonomies.     

A Review of Process Aspects and Modeling of Ebullated Bed Reactors for Hydrocracking of Heavy Oils

The main characteristics of ebullated bed reactors have been reviewed in this work. Key factors of the application of these reactors to hydrocracking of heavy petroleum fractions, such as sediments formation, catalyst attrition and catalyst deactivation, have been clearly discussed. Mathematical representation of ebullated bed systems has been organized into hydrodynamics, scaling down and reactor modeling. Only a few reports dealing with the topic of this review were found in the literature, which employ different levels of sophistication to establish the model equations. These literature reports were summarized and properly discussed, from which it has been recognized that modeling of ebullated bed reactors is a complex task and deserves more attention.     

EXTRACTION OF LIGHT LANTHANIDE NITRATES BY TRI-n-BUTYL PHOSPHATE

The distribution of La, Pr, Nd, Eu, Gd and Tb nitrates between water and tributyl phosphate was investigated over a broad range of concentrations. It has been found that the distribution constant varies only slightly with the lanthanide atomic number, whereas the activity coefficients of the salts in concentrated organic solutions change considerably with both the lanthanide atomic number and concentration. Distribution ratios and separation factors for a number of binary mixtures were determined as a function of the composition at constant total concentration in each phase.     

SOLVENT EXTRACTION OF LANTHANIDES WITH N-m-AND N-p-METHOXYBENZOYL-N-PHENYLHYDROXYLAMINE

The solvent extraction of lanthanides with N-o-methoxy-(o-MBPHA), N-m-methoxy- (m-MBPHA) and N-p-methoxybenzoyl-N-phenylhydroxylamine (p-MBPHA), respectively, in chloroform was studied. The selected lanthanides (Yb, Ho, Eu, Pr and La) were all found extract with m- and p-MBPHA as self-adducts of the form LnL-2HL, where L and HL denote the ligand anion and neutral ligand, respectively. The extraction constant and separation factors for the lanthanides with m- and p-MBPHA were evaluated.     

THIRD PHASE FORMATION IN EXTRACTION OF THORIUM NITRATE BY MIXTURES OF TRIALKYL PHOSPHATES

ABSTRACT

This paper reports the results of the studies on third phase formation during the extraction of thorium nitrate from zero free acidity solutions by mixtures of trialkyl phosphates. The phosphates used are tri n-butyl phosphate(TBP), triiso butyl phosphate(TiBP), tri sec butyl phosphate (TsBP) and tri n-amyl phosphate(TAP). The results indicate that small additions of a homologous phosphate can alter the Limiting Organic Concentration (LOC) above which the third phase formation takes place and thus can be advantageously utilised. Use of mixtures of the trialkyl phosphates as extractant can thus obviate the need for adding modifiers such as alcohols to the organic phase for avoiding third phase formation.