A Meta-analysis on Classification Model Performance in Real-World Datasets: An Exploratory View

The No Free Lunch (NFL) Theorem imposes a theoretical restriction on optimization algorithms and their equal average performance on different problems, under some particular assumptions. Nevertheless, when brought into practice, a perceived “ranking” on the performance is usually perceived by engineers developing machine learning applications. Questions that naturally arise are what kinds of biases the real world has and in which ways can we take advantage from them. Using exploratory data analysis (EDA) on classification examples, we gather insight on some traits that set apart algorithms, datasets and evaluation measures and to what extent the NFL theorem, a theoretical result, applies under typical real-world constraints.     

A deterministic model selection scheme for incremental RBFNN construction in time series forecasting

This paper presents a fast and new deterministic model selection methodology for incremental radial basis function neural network (RBFNN) construction in time series prediction problems. The development of such special designed methodology is motivated by the problems that arise when using a K-fold cross-validation-based model selection methodology for this paradigm: its random nature and the subjective decision for a proper value of K, resulting in large bias for low values and high variance and computational cost for high values. Taking into account these drawbacks, the proposed model selection approach is a combined algorithm that takes advantage of two balanced and representative training and validation sets for their use in RBFNN initialization, optimization and network model evaluation. This way, the model prediction accuracy is improved, getting small variance and bias, reducing the computation time spent in selecting the model and avoiding random and computationally expensive model selection methodologies based on K-fold cross-validation procedures.     

Rheological performance of pectin-enriched products isolated from red beet (Beta vulgaris L. var. conditiva) through alkaline and enzymatic treatments

The rheological performance of pectin-enriched products extracted from red beet (Beta vulgaris L. var. conditiva) root by-products was evaluated in the present work. They were extracted through an alkaline pre-treatment with or without a subsequent enzymatic (hemicellulase or cellulase) hydrolysis at pH 5.2. Flow assays performed with 2.00% w/v-pectin aqueous systems showed pseudoplastic (flow index, n ≈ 0.4 or 0.8) or Newtonian (n = 1.0) behaviour after fitting of experimental data to Ostwald’s law, also showing poor thickening effect. When Ca²⁺ was added to water with the same pectin concentration, true gels developed as confirmed by the mechanical spectra obtained through dynamic assays. Junction zones of homogalacturonan (HG) side chains mediated by Ca²⁺ were able to build up rigid networks in water.Isolated pectins (2.00% w/v) were also used to constitute milk model systems. Whole and skimmed milk were used at two different concentrations. Milk systems showed more transient and weaker gel networks when compared to Ca²⁺-aqueous systems, and were associated to the formation of a [κ-casein?calcium cross linked low methoxyl pectin] complex dampened by the included milk fat globules. Relaxation spectra of pectin-milk systems were in general extended to large relaxation times (10⁴ s) for all isolated fractions studied, which is typical of structured systems. Since all pectin fractions showed very similar chemical composition and molecular weight (average value and distribution), it was suggested that some differences in the rheological performance of each pectin product came from the different length of arabinans and distribution of rhamnose kinks (RG-I, random coil) as well as from the length of demethylated HG chains (semi-flexible coils).The results of this research show that the pectin-enriched fractions isolated from red beet root wastes are useful as additives in food formulation.     

Ni/Fe electrodes prepared by electrodeposition method over different substrates for oxygen evolution reaction in alkaline medium

A series of Ni/Fe electrodes have been prepared by electrodeposition of metal salt precursors on different substrates. The surface morphology, chemical composition and electrochemical characteristics of these electrodes were studied by various physico-chemical techniques such as X-ray Photoelectron Spectroscopy (XPS) and Scanning Electron Microscopy (SEM). The electrochemical properties of the electrodes were examined by steady-state polarization curves. First, the influence of features such as Ni/Fe composition and type of substrate for the oxygen evolution reaction (OER) were determined by electrochemical techniques in a conventional 3-electrodes cell. The overpotential for the OER is lower for the electrodes with the higher concentrations of Ni. The electrodes with a Ni/Fe composition of 75/25 wt.% electrodeposited on steel mesh and/or 75/25 and 50/50 wt.% on nickel foam result in the most active configurations for the OER. These electrodes were further tested as anodes for alkaline water electrolysis during at least 70 h. In order to understand their activity and stability, the used electrodes were also characterized by SEM and compared to the fresh electrodes. Among the compositions and substrates examined, the Ni50Fe50-Nf electrode exhibited the lowest overpotential (2.1 V) for the OER and the higher stability as anode in an alkaline water electrolysis cell.     

Using Combined Bipolar Semantic Scales and Eye‐Tracking Metrics to Compare Consumer Perception of Real and Virtual Bottles

Three‐dimensional virtual representations of consumer products are expected to gain relevance in e‐commerce applications as low cost virtual reality headsets arrive on the market in the next years. However, there are a limited number of studies related to the perceptual evaluation of virtual products and their packaging where virtual and real (photographic) representations are compared. As part of an extensive exploration toward understanding product perception in virtual stores, this work presents a study with 38 participants in which consumer perceptions of a photographic and a virtual representation of a beer bottle are examined. Perceptual evaluation is assessed using two metrics: first, an evaluation was performed by applying a bipolar semantic scale based on four axes: novelty, resolution, style and emotion. Second, eye‐tracking metrics were employed to analyse participant gaze behaviour during the visualization of stimuli. Virtual bottles were modelled using a medium polygonal load (5 K polygons per bottle), and render quality was also medium to intentionally recreate the computing limitations of smartphone‐based virtual reality headsets. Results show that a medium render quality alters consumer perception and responses using semantic scales. Eye‐tracking analysis confirms that the orientation of the bottle and how it is presented also affect consumer perception. While some orientations result in similar eye‐tracking metrics, others show different results. Copyright © 2015 John Wiley & Sons, Ltd.     

Reinforcement of polypropylene with lignocellulose nanofibrils and compatibilization with biobased polymers

Freeze‐dried and milled lignocellulose nanofibrils (LCNF) were used to reinforce polypropylene (PP) nanocomposites. The LCNF, containing up to 9% lignin, was obtained from residual Empty Palm Fruit Bunch (EPFB) fibers. Soy protein isolate (SPI) and hydroxypropyl cellulose (HPC) were tested as coupling agents as well as maleic anhydride grafted polypropylene (MAPP), which was used as a reference. A good level of dispersion of LCNF in the PP matrix while mechanical testing and thermal analyses indicated an improvement of the thermo‐mechanical behavior of the nanocomposites was revealed upon loading of the lignin‐containing nanofibrils. The tensile modulus of PP was increased by 15% upon the addition of 1% LCNF with SPI as a compatibilizer. Likewise, the thermal stability of the composites was most markedly enhanced. Overall, LCNF and SPI, two important bioresources, are introduced here for the development of novel and cost‐effective PP‐based composites. © 2016 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2016 , 133, 43854.     

Senescent CD4+CD28− T Lymphocytes as a Potential Driver of Th17/Treg Imbalance and Alveolar Bone Resorption during Periodontitis

Senescent cells express a senescence-associated secretory phenotype (SASP) with a pro-inflammatory bias, which contributes to the chronicity of inflammation. During chronic inflammatory diseases, infiltrating CD4⁺ T lymphocytes can undergo cellular senescence and arrest the surface expression of CD28, have a response biased towards T-helper type-17 (Th17) of immunity, and show a remarkable ability to induce osteoclastogenesis. As a cellular counterpart, T regulatory lymphocytes (Tregs) can also undergo cellular senescence, and CD28⁻ Tregs are able to express an SASP secretome, thus severely altering their immunosuppressive capacities. During periodontitis, the persistent microbial challenge and chronic inflammation favor the induction of cellular senescence. Therefore, senescence of Th17 and Treg lymphocytes could contribute to Th17/Treg imbalance and favor the tooth-supporting alveolar bone loss characteristic of the disease. In the present review, we describe the concept of cellular senescence; particularly, the one produced during chronic inflammation and persistent microbial antigen challenge. In addition, we detail the different markers used to identify senescent cells, proposing those specific to senescent T lymphocytes that can be used for periodontal research purposes. Finally, we discuss the existing literature that allows us to suggest the potential pathogenic role of senescent CD4⁺CD28⁻ T lymphocytes in periodontitis.