Semantic Explanation for Deep Neural Networks Using Feature Interactions

Given the promising results obtained by deep-learning techniques in multimedia analysis, the explainability of predictions made by networks has become important in practical applications. We present a method to generate semantic and quantitative explanations that are easily interpretable by humans. The previous work to obtain such explanations has focused on the contributions of each feature, taking their sum to be the prediction result for a target variable; the lack of discriminative power due to this simple additive formulation led to low explanatory performance. Our method considers not only individual features but also their interactions, for a more detailed interpretation of the decisions made by networks. The algorithm is based on the factorization machine, a prediction method that calculates factor vectors for each feature. We conducted experiments on multiple datasets with different models to validate our method, achieving higher performance than the previous work. We show that including interactions not only generates explanations but also makes them richer and is able to convey more information. We show examples of produced explanations in a simple visual format and verify that they are easily interpretable and plausible.

RESEARCH OF COMPOSITION OF ANTIWEAR ADDITIVES FOR JET FUELS

In the framework of the state import substitution program for the development of a domestic antiwear additive for jet fuels the physicochemical characteristics of the approved foreign antiwear additives of DNK, HITEC 580 and Unicor J were analyzed at the first stage. The analysis revealed that using solvents during the production of HITEC 580 additives and Unicor J improve their consumer properties (low temperature characteristics). At the second stage of the study the composition of currently used additives HITEC 580 and Unicor J was studied by gas chromatography-mass spectrometry. It was shown, that the same active component is used in their composition - dimers of unsaturated fatty acids C18, namely a mixture of dimers of linolenic, linoleic and oleic acids, among which the dimer formed from linoleic and oleic acids prevails. An analysis of the solvents used in the additive composition showed that in the production of Unicor J additives they use an aromatic solvent containing 92% of monoaromatic compounds, mainly of the C10 series, which is probably the product of gasoline reforming, and HITEC 580 additives use an aliphatic solvent containing mainly n alkanes (41%). From the presented results of the study it follows that when developing a domestic antiwear additive formulation, it is necessary to choose compounds containing 1 or 2 carboxyl groups as the active component, and as the solvent either aromatic or aliphatic, depending on the cost and availability on the domestic market.

Physicochemical Characterization of Packaging Foils Coated by Chitosan and Polyphenols Colloidal Formulations

In this research, antimicrobial polysaccharide chitosan was used as a surface coating for packaging material. The aim of our research was to establish an additive formulation of chitosan and antioxidative plant extracts as dispersion of nanoparticles. Chitosan nanoparticles with embedded thyme, rosemary and cinnamon extracts were synthesized, and characterized for this purpose. Two representative, commercially used foils, polypropylene (PP) and polyethylene (PE), previously activated by UV/ozone to improve coating adhesion, were functionalized using chitosan-extracts nanoparticle dispersions. The foils were coated by two layers. A solution of macromolecular chitosan was applied onto foils as a first layer, followed by the deposition of various extracts embedded into chitosan nanoparticles that were attached as an upper layer. Since active packaging must assure bioactive efficiency at the interface with food, it is extremely important to understand the surface characteristics and phenomena of functionalized foils. The physico-chemical analyses of functionalized foils were thus comprised of surface elemental composition, surface charge, wettability, as well as surface morphology. It has been shown that coatings were applied successfully with an elemental composition, surface charge and morphology that should enable coating stability, homogeneity and consequently provide an active concept of the packaging surface in contact with food. Moreover, the wettability of foils was improved in order to minimize the anti-fogging behavior.

Effects of hydraulic oil and lubricant additives on dynamic friction properties under various reciprocating sliding conditions

The friction characteristics of a shock absorber are very complex because the reciprocating motion is not always identical. In this study a device was developed and used to analyze the dynamic friction characteristics under various reciprocating sliding conditions to determine the sliding materials and hydraulic oils that improve the shock absorber performance. This study describes the influence of hydraulic oil additive on the fine reciprocating friction characteristics of steel and copper alloy. Hydraulic oils were prepared by blending a paraffinic mineral oil with zinc dithiophosphate (ZnDTP) and polyhydric alcohol ester as additives. The results show that the dynamic frictional characteristics vary mainly depending on the additive concentration. A specific additive formulation induces a unique amplitude-dependent friction behavior. In addition, the influence of different additives on the lubrication mechanism is investigated based on the instrumental analysis of the friction surface.

FDM 3D Printing of Polymers Containing Natural Fillers: A Review of their Mechanical Properties

As biodegradable thermoplastics are more and more penetrating the market of filaments for fused deposition modeling (FDM) 3D printing, fillers in the form of natural fibers are convenient: They have the clear advantage of reducing cost, yet retaining the filament biodegradability characteristics. In plastics that are processed through standard techniques (e.g., extrusion or injection molding), natural fibers have a mild reinforcing function, improving stiffness and strength, it is thus interesting to evaluate whether the same holds true also in the case of FDM produced components. The results analyzed in this review show that the mechanical properties of the most common materials, i.e., acrylonitrile-butadiene-styrene (ABS) and PLA, do not benefit from biofillers, while other less widely used polymers, such as the polyolefins, are found to become more performant. Much research has been devoted to studying the effect of additive formulation and processing parameters on the mechanical properties of biofilled 3D printed specimens. The results look promising due to the relevant number of articles published in this field in the last few years. This notwithstanding, not all aspects have been explored and more could potentially be obtained through modifications of the usual FDM techniques and the devices that have been used so far.

Lithium-Ion Capacitors and Hybrid Lithium-Ion Capacitors—Evaluation of Electrolyte Additives Under High Temperature Stress

ABSTRACTLithium-ion capacitors (LICs) and Hybrid LICs (H-LICs) were assembled as three-layered pouch cells in an asymmetric configuration employing Faradaic pre-lithiated hard carbon anodes and non-Faradaic ion adsorption-desorption activated carbon (AC) cathodes for LICs and lithium iron phosphate (LiFePO4-LFP)/AC composite cathodes for H-LICs. The room temperature rate performance was evaluated after the initial LIC and H-LIC cell formation as a function of the electrolyte additives. The capacity retention was measured after charging at high temperature conditions, while the design factor explored was electrolyte additive formulation, with a focus on their stability. The high temperature potential holds simulate electrochemical energy materials under extreme environments and act to accelerate the failure mechanisms associated with cell degradation to determine robust electrolyte/additive combinations.

Active Contours Using Additive Local and Global Intensity Fitting Models for Intensity Inhomogeneous Image Segmentation

This paper introduces an improved region based active contour method with a level set formulation. The proposed energy functional integrates both local and global intensity fitting terms in an additive formulation. Local intensity fitting term influences local force to pull the contour and confine it to object boundaries. In turn, the global intensity fitting term drives the movement of contour at a distance from the object boundaries. The global intensity term is based on the global division algorithm, which can better capture intensity information of an image than Chan-Vese (CV) model. Both local and global terms are mutually assimilated to construct an energy function based on a level set formulation to segment images with intensity inhomogeneity. Experimental results show that the proposed method performs better both qualitatively and quantitatively compared to other state-of-the-art-methods.

pH dependent supramolecular recognition of dapoxyl sodium sulfonate with 2-hydroxypropyl β-cyclodextrin: an application towards food-additive formulation

pH dependent host–guest complexation of dapoxyl sodium sulfonate (DSS), an intramolecular charge transfer dye, with 2-hydroxypropyl beta-cyclodextrin (HP-β-CD) has been investigated.