Disentanglement Mechanism Research Articles

Beyond "Taming Electric Scooters": Disentangling Understandings of Micromobility Naturalistic Riding

Electric(e)-scooters have emerged as a popular, ubiquitous, and first/last-mile micromobility transportation option within and across many cities worldwide. With the increasing situation-awareness and on-board computational capability, such intelligent micromobility has become a critical means of understanding the rider's interactions with other traffic constituents (called Rider-to-X Interactions, RXIs), such as pedestrians, cars, and other micromobility vehicles, as well as road environments, including curbs, road infrastructures, and traffic signs. How to interpret these complex, dynamic, and context-dependent RXIs, particularly for the rider-centric understandings across different data modalities --- such as visual, behavioral, and textual data --- is essential for enabling safer and more comfortable micromobility riding experience and the greater good of urban transportation networks. Under a naturalistic riding setting (i.e., without any unnatural constraint on rider's decision-making and maneuvering), we have designed, implemented, and evaluated a pilot Cross-modality E-scooter Naturalistic Riding Understanding System, namely CENRUS, from a human-centered AI perspective. We have conducted an extensive study with CENRUS in sensing, analyzing, and understanding the behavioral, visual, and textual annotation data of RXIs during naturalistic riding. We have also designed a novel, efficient, and usable disentanglement mechanism to conceptualize and understand the e-scooter naturalistic riding processes, and conducted extensive human-centered AI model studies. We have performed multiple downstream tasks enabled by the core model within CENRUS to derive the human-centered AI understandings and insights of complex RXIs, showcasing such downstream tasks as efficient information retrieval and scene understanding. CENRUS can serve as a foundational system for safe and easy-to-use micromobility rider assistance as well as accountable use of micromobility vehicles.

Anonymizing medical case-based explanations through disentanglement

Case-based explanations are an intuitive method to gain insight into the decision-making process of deep learning models in clinical contexts. However, medical images cannot be shared as explanations due to privacy concerns. To address this problem, we propose a novel method for disentangling identity and medical characteristics of images and apply it to anonymize medical images. The disentanglement mechanism replaces some feature vectors in an image while ensuring that the remaining features are preserved, obtaining independent feature vectors that encode the images’ identity and medical characteristics. We also propose a model to manufacture synthetic privacy-preserving identities to replace the original image’s identity and achieve anonymization. The models are applied to medical and biometric datasets, demonstrating their capacity to generate realistic-looking anonymized images that preserve their original medical content. Additionally, the experiments show the network’s inherent capacity to generate counterfactual images through the replacement of medical features.

DDOD: Dive Deeper into the Disentanglement of Object Detector

Compared to many other dense prediction tasks, object detection plays a fundamental role in visual perception and scene understanding. Dense object detection, aiming at localizing objects directly from the feature map, has drawn great attention due to its low cost and high efficiency. Though it has been developed for a long time, the training pipeline of dense object detectors is still compromised to lots of conjunctions. In this paper, we demonstrate the existence of three conjunctions lying in the current paradigm of one-stage detectors: 1) only samples assigned as positive in classification head are used to train the regression head; 2) classification and regression share the same input feature and computational fields defined by the parallel head architecture; and 3) samples distributed in different feature pyramid layers are treated equally when computing the loss. Based on this, we propose Disentangled Dense Object Detector (DDOD), a simple, direct, and efficient framework for 2D detection with strong performance. We derive two DDOD variants (i.e., DR-CNN, and DDETR) following the basic one-stage/two-stage and recently developed transformer-based pipelines. Specifically, we develop three effective disentanglement mechanisms and integrate them into the current state-of-the-art object detectors. Extensive experiments on MS COCO benchmark show that our approach obtains significant enhancements with negligible extra overhead on various detectors. Notably, our best model reaches 55.4 mAP on the COCO <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">test-dev</i> set, achieving new state-of-the-art performance on this competitive benchmark. Additionally, we validate our model on several challenging tasks including small object detection and crowded object detection. The experimental results further prove the superiority of disentanglement on these conjunctions. Code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/zehuichen123/DDOD</uri> .

Disentangled Prototypical Graph Convolutional Network for Phishing Scam Detection in Cryptocurrency Transactions

Blockchain technology has generated an influx of transaction data and complex interactions, posing significant challenges for traditional machine learning methods, which struggle to capture high-dimensional patterns in transaction networks. In this paper, we present the disentangled prototypical graph convolutional network (DP-GCN), an innovative approach to account classification in Ethereum transaction records. Our method employs a unique disentanglement mechanism that isolates relevant features, enhancing pattern recognition within the network. Additionally, we apply prototyping to disentangled representations, to classify scam nodes robustly, despite extreme class imbalances. We further employ a joint learning strategy, combining triplet loss and prototypical loss with a gamma coefficient, achieving an effective balance between the two. Experiments on real Ethereum data showcase the success of our approach, as the DP-GCN attained an F1 score improvement of 32.54%p over the previous best-performing GCN model and an area under the ROC curve (AUC) improvement of 4.28%p by incorporating our novel disentangled prototyping concept. Our research highlights the importance of advanced techniques in detecting malicious activities within large-scale real-world cryptocurrency transactions.

StyleTerrain: A novel disentangled generative model for controllable high-quality procedural terrain generation

Terrain is a vital element in construction of virtual scene in the digital era. Despite considerable progress has been made in Generative Adversarial Network (GAN) based terrain modeling methods, their quality and controllability still cannot meet up the requirements of many emerging industries. The present work proposes a novel disentangled generative model, named as StyleTerrain, for achieving controllable high-quality terrain generation. It introduces disentangled representation learning into GAN-based terrain modeling methods for the first time. The model has been evaluated quantitatively. The results show a significantly short perceptual path length and the effectiveness of the disentanglement mechanism in controllable terrain generation, indicating that latent space disentanglement is a promising future direction for achieving generation controllability in GAN-based terrain modeling methods.

Size effects of silica on the viscosity of ultra‐high molecular weight polyethylene and its disentanglement mechanism

AbstractThe viscosity of the polymer composites is closely related to the size effect of the filler. In this study, the size effects of silica fillers on the viscosity of ultrahigh molecular weight polyethylene (UHMWPE) was investigated. The addition of 200 nm and 5 μm silica increase the viscosity, but the 7 nm silica can significantly decrease the viscosity. In order to understand the mechanism of this phenomenon, the multilevel factors of the sample in terms of matrix structure (amorphous phase entanglement, mesophase and crystalline region content) and the mobility of chain segments were analyzed. Rheology and DMA tests have shown that silica affects viscosity mainly by influencing the entanglement in the amorphous phase. Raman tests show that 7 nm silica decrease the viscosity mainly by affecting the content of the mesophase and the mobility of the chain segments in the amorphous region. The annealing process was investigated by Flash DSC, and it was found that 7 nm SiO2 can accelerate the mobility of UHMWPE chains to form a larger recrystallization area, and the addition of 200 nm and 5 μm SiO2 did not show a similar phenomenon, proving the strong size effect of SiO2 on the viscosity and disentanglement of UHMWPE.

You Only Look Yourself: Unsupervised and Untrained Single Image Dehazing Neural Network

In this paper, we study two challenging and less-touched problems in single image dehazing, namely, how to make deep learning achieve image dehazing without training on the ground-truth clean image (unsupervised) and an image collection (untrained). An unsupervised model will avoid the intensive labor of collecting hazy-clean image pairs, and an untrained model is a “real” single image dehazing approach which could remove haze based on the observed hazy image only and no extra images are used. Motivated by the layer disentanglement, we propose a novel method, called you only look yourself (YOLY) which could be one of the first unsupervised and untrained neural networks for image dehazing. In brief, YOLY employs three joint subnetworks to separate the observed hazy image into several latent layers, i.e., scene radiance layer, transmission map layer, and atmospheric light layer. After that, three layers are further composed to the hazy image in a self-supervised manner. Thanks to the unsupervised and untrained characteristics of YOLY, our method bypasses the conventional training paradigm of deep models on hazy-clean pairs or a large scale dataset, thus avoids the labor-intensive data collection and the domain shift issue. Besides, our method also provides an effective learning-based haze transfer solution thanks to its layer disentanglement mechanism. Extensive experiments show the promising performance of our method in image dehazing compared with 14 methods on six databases. The code could be accessed at www.pengxi.me.

Shear Thinning: Determination of Zero‐Shear Viscosities from Measurements in the Non‐Newtonian Region

AbstractExperimental information on the viscosities, η, of polymer solutions and of polymer melts as a function of shear rate is modeled by means of an approach that describes the diminution of ln η as a function of shear stress, τ, in terms of an exponential decay. The approach uses the following three adjustable parameters: the zero‐shear viscosity of the system, a characteristic shear stress, quantifying its susceptibility toward shear thinning, and a dimensionless parameter stating the magnitude of the effect. This procedure gives access to the Newtonian behavior also in cases where direct measurements are impractical or impossible; it discloses two phenomena not reported so far: a qualitative change in the efficacy of τ at a characteristic concentration and indicates the occurrence of two different disentanglement mechanisms in thermodynamically unfavorable solvents.

Short‐time fabrication of well‐mixed high‐density polyethylene/ultrahigh‐molecular‐weight polyethylene blends under elongational flow: morphology, mechanical properties and mechanism

AbstractAs linear polyethylenes, ultrahigh‐molecular‐weight polyethylene (UHMWPE) and high‐density polyethylene (HDPE) have the same molecular structure, but the large difference in viscosity between them makes it difficult to obtain well‐mixed blends. An innovative eccentric rotor extruder (ERE) generating an elongational flow was used to prepare HDPE/UHMWPE blends within short processing times. Compared with the obvious two‐phase morphology of a sample from a twin‐screw extruder observed with a scanning electron microscope, few small UHMWPE particles were observed in the HDPE matrix for a sample from the ERE, indicating the good mixing on a molecular level of HDPE/UHMWPE blends achieved by the ERE during short processing times. The morphological changes of blends prepared using the ERE evidenced the good integration of HDPE and UHMWPE even though the UHMWPE content is up to 50 wt% in the blends. Moreover, all blends retained most of the intrinsic molecular weight. The good mixing was further confirmed from the thermal, crystallization and rheological behaviors determined using differential scanning calorimetry and dynamic rheological measurements. Importantly, the 50/50 blend presented improved mechanical properties, especially super‐impact strength of 151.9 kJ m−2 with incomplete‐break fracture state. The strengthening and great toughening effects of UHMWPE on the blends were attributed to the addition of unwrapped UHMWPE long molecular chains. The effective disentanglement mechanism of UHMWPE chains under elongational flow was explained schematically by a non‐parallel three‐plate model. © 2019 Society of Chemical Industry

Molecular mechanisms of interfacial slip for polymer melts under shear flow

Interfacial slip plays a crucial role in a variety of fluid dynamics problems occurring in practical polymer processing, lubrication, adhesion, nanocomposites, etc. Despite many research efforts, a fundamental understanding of the underlying molecular mechanisms and dynamics is still lacking. Here, we present the intrinsic molecular characteristics of the slip phenomena by using atomistic nonequilibrium molecular dynamics simulations of polyethylene melts under shear flow. Our results identify three distinctive characteristic regimes with regard to the degree of slip and reveal the underlying molecular mechanisms for each regime: (i) the z-to-x chain rotation mechanism in the vorticity plane in the weak flow regime, which effectively diminishes the wall friction against chain movement along the flow direction, (ii) the repetitive chain detachment-attachment (out-of-plane wagging) and disentanglement mechanism in the intermediate regime, and (iii) irregular (chaotic) chain rotation and tumbling mechanisms ...

Wall Slip of Tridisperse Polymer Melts and the Effect of Unentangled versus Weakly Entangled Chains

We characterized wall slip of tridisperse linear 1,4-polybutadiene on a silicon wafer in a parallel plate shear cell and tracer particle velocimetry. Tridisperse mixtures of fixed weight-average molecular weight Mw and varying number-average molecular weight Mn were prepared from nearly monodisperse polybutadienes. Their steady state slip behavior was examined at shear rates over the range of ∼0.1–15 s–1. The results show that the slip behavior in the transition regime depends on Mn at constant Mw. This study also revealed that weakly entangled and unentangled chains in the mixtures influence wall slip differently: mixtures containing moderate amounts of weakly entangled chains exhibited enhanced slip while those containing unentangled chains did not. We explain this observation using the tube theory through the slip disentanglement mechanism proposed by Brochard and de Gennes in 1992 and conclude that the slip behavior is changed because of the force balance between the mobile and adsorbed chains and the...

Effect of aging on the stress crack resistance of an HDPE geomembrane

The changes in the stress crack resistance, SCR (measured by the single point notched constant tensile load; ASTM D5397 – appendix) for a 1.0 mm high-density polyethylene (HDPE) geomembrane aged in synthetic leachate at six temperatures (25, 40, 55, 70, 85 and 95 °C) and in air and water at 55 °C were investigated for almost five years. There are observed changes in the SCR before the geomembrane chemically degraded (as manifest by melt index and tensile properties) and even before the antioxidants depletion at temperatures below 70 °C (as manifest by the standard and high pressure oxidative induction times). This change is attributed to morphological changes during aging that affected the inter-lamellar connections due to: (a) annealing that increased the strength of the inter-lamellar connections (over the temperature ranges examined), and (b) the proposed chain disentanglement mechanism. The annealing and chain disentanglement mechanisms have a counteracting effect on the inter-lamellar connections and hence on the SCR. Chain disentanglement dominated over the annealing effect at 25, 40, 55 and 70 °C, resulting, respectively, in a decrease in SCR to an equilibrium SCRm of 0.83, 0.65, 0.26 and 0.43 of the initial stress crack resistance of the geomembrane (SCRo). The maximum decrease in SCR is observed when the geomembrane is aged in leachate at 55 °C with the SCR decreasing to 0.26 SCRo. The SCR of the geomembrane aged in air and water at 55 °C decreased to about 0.5 SCRo, implying that the surfactant used in leachate may have enhanced the rate and the extent of chain disentanglement (i.e., decreasing the time to reach a lower SCRm) due to plasticization of the amorphous zone.

Experimental study and molecular dynamics simulation of wall slip in a micro-extrusion flowing process

Wall slip is one of the most important characteristics of polymer melts’ elasticity behaviours as well as the most significant factor which affects the flow of polymer melts. Based on the traditional Mooney method, through a double-barrel capillary rheometer, the relationship between velocities of wall slip, shear stress, shear rate, diameters of dies, and temperature of polypropylene (PP), high-density polyethylene (HDPE), polystyrene (PS), and polymethylmethacrylate (PMMA) is explored. The results indicate that the velocities of the wall slip of PP and HDPE increase apparently with shear stress and slightly with temperature. Meanwhile, the rise of temperature results in the decrease of critical shear stress. The wall-slip velocities of PS and PMMA are negative which means that the Mooney method based on the adsorption–desorption mechanism has determinate limitation to calculate the wall-slip velocity. Based on the entanglement–disentanglement mechanism, a new wall-slip model is built. With the new model, the calculation values of velocity of PP and HDPE correspond to the experimental values very well and the velocities of PS and PMMA are positive. The velocities of PS and PMMA increase obviously with the rise of shear stress. The rise of temperature results in the increase of velocity and decrease of critical shear stress. Then, the molecular dynamics simulation is used to investigate the combining energy between four polymer melts and the inside wall. The results show that at the given temperature and pressure, the molecules of PS and PMMA combine with atoms of the wall more tightly than those of PP and HDPE which means when wall slip occurs, the molecules of PS and PMMA near the wall will adsorb to the surface of the wall. However, those of PP and HDPE will be easy to slip. Therefore, the wall-slip mechanism of PP and HDPE is the adsorption–desorption mechanism, and that of PS and PMMA is the entanglement–disentanglement mechanism. According to the different wall-slip mechanisms of four polymers, an all-sided calculation method of wall-slip velocity is raised which consummates the theory of wall slip of polymer melts.

Entanglement‐buildup through charged‐exciton decay in a semiconductor quantum dot

We analyze the decay of a single-electron charged exciton in a quantum dot embedded in a field effect structure. We show how the quantum properties of the charged exciton are transferred through tunneling and relaxation to the spin entanglement between electrons in the dot and contact, carefully examine the proper theoretical description of the underlying scattering processes, and identify the pertinent disentanglement mechanisms. (© 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)

Single scatterings in single artificial atoms: Quantum coherence and entanglement

We employ the quantum-jump approach to study single scatterings in single semiconductor quantum dots. Two prototypical situations are investigated. First, we analyze two-photon emissions from the cascade biexciton decay of a dot where the single-exciton states exhibit a fine-structure splitting. We show that this splitting results for appropriately chosen polarization filters in an oscillatory behavior of two-photon correlations, and carefully examine the proper theoretical description of the underlying scattering processes. Secondly, we analyze the decay of a single-electron charged exciton in a quantum dot embedded in a field effect structure. We show how the quantum properties of the charged exciton are transferred through tunneling and relaxation to the spin entanglement between electrons in the dot and contact, and identify the pertinent disentanglement mechanisms.

A turnstile electron-spin entangler in semiconductors

We propose a single-electron doped quantum dot in a field-effect structure as an optically triggered turnstile for spin-entangled electrons. A short laser pulse excites a charged exciton, whose quantum properties are transferred through tunneling and relaxation to the spin entanglement between electrons in the dot and contact. We identify the pertinent disentanglement mechanisms, and discuss experimental detection and possible application schemes.

Studies on Wall-Slip in Entangled Polymeric Liquids

Abstract In this work, it is attempted to theoretically understand the phenomenon of wall slip through empirical and molecular models. Initially, we use the framework a transient network theory. We show that the severe disentanglement in the interfacial region can give rise to non-monotonic flow curve locally in that region. Further, we generalize this model into a unified slip model, which predicts wall slip by either disentanglement or by debonding mechanism, depending upon the adhesive energy of the wall-polymer pair. The model predictions of the critical wall shear stress are in good agreement with experiments for various adhesive energies of the wall-polymer pair. The model predicts that the temperature dependence of the critical wall shear stress for debonding is different than that of disentanglement mechanism under certain experimental conditions. To validate the predictions of unified model, we measure the critical stress for sudden slip due to debonding for various temperatures using cone and plate viscometer with fluoroelastomer-coated cone. The temperature dependence of the critical stress for instability (slip) on a coated cone is found out to be inversely dependent on temperature, which expected for the case of debonding. In the final part of this thesis, we develop a parameter-free tube model for predicting the stick-slip phenomenon. The model, which is based on the contour variable model [Mead et al., 1998, Macromolecules, 31, 7895], considers the dynamics of the tethered chains, which are grafted on a high-energy wall and which are entangled with the bulk chains flowing past them. We show that the restricted relaxation modes of the tethered molecule give rise to discontinuous slip instability. More specifically, the slow relaxation of the tethered chain due to the restricted convective constraint release is unable to randomize its flow-induced orientation above a critical shear rate or stress. This decreases the resistance to flow for the bulk chains, which suddenly slip past the oriented tethered chains.

A unified wall slip model

A unified slip model is developed, which predicts wall slip by either a disentanglement mechanism or by debonding mechanism, depending upon the adhesive energy of the wall-polymer pair. The model is based on the transient network theory, in which the activation processes of adsorption and desorption are considered to occur at the wall in parallel to the stretching of the adsorbed chains. It is shown that the stick–slip transition occurs due to the local non-monotonic flow behavior near the wall irrespective of the mechanism of slip. The model predictions of the critical wall shear stress are in good agreement with experimentally observed values of the critical stress for various adhesive energies of wall polymer pair. Another important prediction of the model is that the temperature dependence of the critical wall shear stress for debonding is different than that of disentanglement mechanism under certain experimental conditions. This may be useful for discerning the correct mechanism of slip. The unified model encompasses different systems (viz. entangled solutions and melts) and diverse mechanisms (viz. disentanglement and debonding) in a common mathematical framework.

Effect of Film Thickness on Rheological Behavior of Asphalt Binders

In the current Superpave® binder tests, the rheological properties of asphalt binders are determined by the dynamic shear rheometer. The asphalt binders are tested using a gap size of either 1 or 2 mm. In the asphalt mixture, the thickness of the asphalt film between aggregates can vary between a few microns and a few millimeters. Because of the interaction between the asphalt and the aggregate, rheological behavior in thin films can differ from the behavior determined at large gap sizes. In this study the rheological properties of asphalt binders were measured at different gap sizes that simulate the range in film thickness of a typical mixture. Significant changes in rheological properties were observed as the gap size changed. The results indicate that different modified binders show different trends as the gap size changes. A review of the literature was conducted for an explanation of this behavior. Four mechanisms were offered to explain similar gap effects observed for other materials. Using the data collected in this study, the disentanglement mechanism caused by adhesion to testing plates appears to fit the observations for asphalt binders. The results suggest that the current gap settings in the binder tests might not be sufficient to determine the performance of the binder in the mixture. The findings also suggest that more work is required to understand the effect on the rheological behavior of asphalt binders of physical and chemical interactions with aggregates.

Self-oscillations in capillary flow of entangled polymers

A well-established “coextrusion” technique is applied to visualize and document a newly found self-oscillation due to an unstable boundary condition in capillary flow of high-density polyethylene. In contrast to the other types of self-oscillations of interfacial origin reported in the literature, the phenomenon can only occur in a die where the upstream portion of the capillary die wall has a surface condition different from that of the downstream portion. The self-oscillation phenomenon takes place under a constant pressure when one half of the die wall (coated with a layer of fluoropolymer) assumes steady slip due to adhesive breakdown, and the boundary condition of the other half switches between no-slip and slip states through the cohesive chain disentanglement mechanism.