Application Research for Fusion Model of Pseudolabel and Cross Network.

Datasets usually suffer from supervised information missing and weak generalization ability in deep convolution neural network. In this paper, pseudolabel (PL) of Weakly Supervised Learning (WSL) was used to address the problem of supervised information missing, while Cross Network (CN) of Multitask Learning (MTL) was used to solve the problem of weak generalization ability in deep convolution neural network. In PL, the data of supervised information missing was predicted; thus, PL of the corresponding data was generated. In CN, PL data and labeled data were taken as two tasks to train together. Firstly, the labeled data was divided into training dataset and testing dataset, respectively, and image preprocessing was carried out. Secondly, the network was initialized and trained, and the model with high accuracy and good generalization was selected as the optimal model. Then, the optimal model was used to predict the unlabeled data and generate PL. Finally, the steps above were repeated several times to find a better optimal model. In the experiments of the fusion model of PL and CN, Facial Beauty Prediction was regarded as main task and the others as auxiliary tasks. Experimental results show that the model was suitable for multitask training of different tasks in different or similar datasets, and the accuracy of the main task of Facial Beauty Prediction reaches 64.76%, higher than the highest accuracy by conventional methods.

Structural design approaches for creating fat droplet and starch granule mimetics.

This article focuses on hydrogel-based strategies for creating reduced calorie foods with desirable physicochemical, sensory, and nutritional properties. Initially, the role of fat droplets and starch granules in foods is discussed, and then different methods for fabricating hydrogel beads are reviewed, including phase separation, antisolvent precipitation, injection, and emulsion template methods. Finally, the potential application of hydrogel beads as fat droplet and starch granule replacements is discussed. There is still a need for large-scale, high-throughout, and economical methods of fabricating hydrogel beads suitable for utilization within the food industry.

Stabilization of biopolymer microgels formed by electrostatic complexation: Influence of enzyme (laccase) cross-linking on pH, thermal, and mechanical stability.

Biopolymer microgels formed by electrostatic complexation are often susceptible to disintegration when environmental conditions are changed, and so methods are required to improve their stability. In this study, microgels were formed by electrostatic complexation of a protein (type-B gelatin) and a polysaccharide (beet pectin). The impact of enzyme (laccase) crosslinking of the ferulic acid groups on the beet pectin was then studied as a method to improve microgel stability to environmental stresses. Gelatin-beet pectin (1:0.25w/w) microgels were formed at 35°C and pH4.4, and then the pH dependence of the ζ-potential, size, turbidity, and microstructure of the microgels was measured in the absence and presence of laccase cross-linking. Our results suggested that crosslinking occurred within the microgels (rather than between them) since no particle aggregation was observed after enzyme treatment. Enzyme crosslinking did not affect the ζ-potential of the microgels, but it did decrease their size. Both cross-linked and non-cross-linked microgels were stable to aggregation at low (2-3) and high (4.4-7) pH values, but not at intermediate values (3-4.4), which was attributed to their low surface charge. Cross-linking improved the resistance of the microgels to shearing-induced disruption (300rpm for 24h) and to thermal-induced disruption (50°C for 2min). These cross-linked biopolymer microgels may have applications for texture modification, encapsulation, or controlled release.

Development of hydrocolloid microgels as starch granule mimetics: Hydrogel particles fabricated from gelatin and pectin.

In this study, hydrocolloid microgels fabricated by electrostatic complexation of gelatin and pectin were developed as possible starch mimetics. The impact of covalent cross-linking on the physicochemical and structural properties of the microgels was investigated. Microgels were formed by acidifying a mixture of gelatin (0.5wt.%) and pectin (0.01wt.%) from pH10 to 5 at 40°C, followed by cross-linking with glutaraldehyde (0 to 2mM). At low glutaraldehyde levels (<0.5mM), cross-linking occurred primarily within the microgels and did not affect particle dimensions, whereas at high levels (2mM), cross-linking connected adjacent microgels leading to the formation of large flocs. Rheological and microscopic analysis showed that the degree of cross-linking impacted the thermal transitions of the microgels. A simulated oral processing study indicated that the melt-in-the-mouth behavior of the hydrocolloid microgels could be made to be similar to that of starch granules by controlling the degree of cross-linking. This study may be useful for designing starch mimetics with improved texture-modifying properties and reduced-calories.

Design of reduced-fat food emulsions: Manipulating microstructure and rheology through controlled aggregation of colloidal particles and biopolymers.

The objective of this study was to develop model reduced-calorie food emulsions with desirable textural and optical properties based on controlled aggregation of food-grade colloidal particles and biopolymers. The model food emulsion consisted of fat droplets (5wt.%), starch granules (4wt.%), and xanthan gum (0 to 0.02wt.%) under acidic conditions (pH3). The fat droplets were stabilized by a protein-based emulsifier (whey protein isolate). Fat droplet aggregation was induced by adding anionic xanthan gum to promote bridging flocculation of the cationic protein-coated fat droplets. Thermal processing (95°C) did not have a major impact on fat droplet aggregation, but it did promote starch granule swelling. The structural organization of the fat droplets could be regulated by altering xanthan levels. Relatively small droplet aggregates were formed at low xanthan concentrations that coated the starch granule surfaces. Conversely, large irregular shaped droplet aggregates were formed throughout the system at higher xanthan levels. The rheological and optical properties of the model emulsions could therefore be controlled by altering fat droplet organization. Addition of low levels of xanthan significantly increased the viscosity, yield stress, and complex modulus of the model food emulsions. However, high levels of xanthan led to the formation of large visible aggregates that would negatively impact on sensory quality. This study has important implications for the development of cost-effective and clean-label reduced-fat products with desirable quality attributes, such as dressings and sauces.

Soft matter strategies for controlling food texture: formation of hydrogel particles by biopolymer complex coacervation.

Soft matter physics principles can be used to address important problems in the food industry. Starch granules are widely used in foods to create desirable textural attributes, but high levels of digestible starch may pose a risk of diabetes. Consequently, there is a need to find healthier replacements for starch granules. The objective of this research was to create hydrogel particles from protein and dietary fiber with similar dimensions and functional attributes as starch granules. Hydrogel particles were formed by mixing gelatin (0.5 wt%) with pectin (0 to 0.2 wt%) at pH values above the isoelectric point of the gelatin (pH 9, 30 °C). When the pH was adjusted to pH 5, the biopolymer mixture spontaneously formed micron-sized particles due to electrostatic attraction of cationic gelatin with anionic pectin through complex coacervation. Differential interference contrast (DIC) microscopy showed that the hydrogel particles were translucent and spheroid, and that their dimensions were determined by pectin concentration. At 0.01 wt% pectin, hydrogel particles with similar dimensions to swollen starch granules (D3,2 ≈ 23 µm) were formed. The resulting hydrogel suspensions had similar appearances to starch pastes and could be made to have similar textural attributes (yield stress and shear viscosity) by adjusting the effective hydrogel particle concentration. These hydrogel particles may therefore be used to improve the texture of reduced-calorie foods and thereby help tackle obesity and diabetes.

Creation of reduced fat foods: influence of calcium-induced droplet aggregation on microstructure and rheology of mixed food dispersions.

The impact of calcium-induced fat droplet aggregation on the microstructure and physicochemical properties of model mixed colloidal dispersions was investigated. These systems consisted of 2 wt% whey protein-coated fat droplets and 4 wt% modified starch granules heated to induce starch swelling (pH 7). Optical and confocal microscopy showed that the fat droplets were dispersed within the interstitial region between the swollen starch granules. The structural organisation of the fat droplets within these interstitial regions could be modulated by controlling the calcium concentration: (i) at a low calcium concentration the droplets were evenly distributed; (ii) at an intermediate calcium concentration they formed a layer around the starch granules; (iii) at a high calcium concentration they formed a network of aggregated droplets. Paste-like materials were produced when the fat droplets formed a three-dimensional network in the interstitial region. The properties of fat droplet-starch granule suspensions can be modulated by altering the electrostatic interactions to alter microstructure.