Learning subtree pattern importance for Weisfeiler-Lehman based graph kernels

Machine Learning - Graph is an usual representation of relational data, which are ubiquitous in many domains such as molecules, biological and social networks. A popular approach to learning with...     

Wall-following control of a two-wheeled mobile robot

Wall-following control problem for a mobile robot is to move it along a wall at a constant speed and keep a specified distance to the wall.This paper proposes wall-following controllers based on Lyapunov function candidate for a two-wheeled mobile robot (MR) to follow an unknown wall. The mobile robot is considered in terms of kinematic model in Cartesian coordinate system. Two wall-following feedback controllers are designed: full state feedback controller and observer-based controller. To design the former controller, the errors of distance and orientation of the mobile robot to the wall are defined, and the feedback controller based on Lyapunov function candidate is designed to guarantee that the errors converge to zero asymptotically. The latter controller is designed based on Busawon’s observer as only the distance error is measured. Additionally, the simulation and experimental results are included to illustrate the effectiveness of the proposed controllers.     

Sagnac-loop phase shifter with polarization-independent operation

A bulk-optics, polarization-independent phase shifter for photonic quantum information applications is demonstrated. The device is based on a transverse electro-optic modulator in a Sagnac-type loop containing Faraday rotators. This direction-dependent polarization rotation in the loop is combined with a Mach-Zehnder interferometer to form an optical switch with an on/off contrast of 96% and a switching time of 1.6 ns. We demonstrate that the device's operation maintains any arbitrary polarization state.     

Temperature dependence dynamical permeability characterization of magnetic thin film using near-field microwave microscopy

A temperature dependence characterization system of microwave permeability of magnetic thin film up to 5 GHz in the temperature range from room temperature up to 423 K is designed and fabricated as a prototype measurement fixture. It is based on the near field microwave microscopy technique (NFMM). The scaling coefficient of the fixture can be determined by (i) calibrating the NFMM with a standard sample whose permeability is known; (ii) by calibrating the NFMM with an established dynamic permeability measurement technique such as shorted microstrip transmission line perturbation method; (iii) adjusting the real part of the complex permeability at low frequency to fit the value of initial permeability. The algorithms for calculating the complex permeability of magnetic thin films are analyzed. A 100 nm thick FeTaN thin film deposited on Si substrate by sputtering method is characterized using the fixture. The room temperature permeability results of the FeTaN film agree well with results obtained from the established short-circuited microstrip perturbation method. Temperature dependence permeability results fit well with the Landau-Lifshitz-Gilbert equation. The temperature dependence of the static magnetic anisotropy H(K)(sta), the dynamic magnetic anisotropy H(K)(dyn), the rotational anisotropy H(rot), together with the effective damping coefficient α(eff), ferromagnetic resonance f(FMR), and frequency linewidth Δf of the thin film are investigated. These temperature dependent magnetic properties of the magnetic thin film are important to the high frequency applications of magnetic devices at high temperatures.     

Deep Appearance Models: A Deep Boltzmann Machine Approach for Face Modeling

International Journal of Computer Vision - The “interpretation through synthesis” approach to analyze face images, particularly Active Appearance Models (AAMs) method, has become one of...     

Learning from Longitudinal Face Demonstration—Where Tractable Deep Modeling Meets Inverse Reinforcement Learning

International Journal of Computer Vision - This paper presents a novel subject-dependent deep aging path (SDAP), which inherits the merits of both generative probabilistic modeling and inverse...     

User identification via neural network based language models

Identifying compromised accounts on online social networks that are used for phishing attacks or sending spam messages is still one of the most challenging problems of cyber security. In this research, the authors explore an artificial neural network‐based language model to differentiate the writing styles of different users on short text messages. In doing so, the aim is to be able to identify compromised user accounts. The results obtained indicate that one can learn the language model on one dataset and can generalize it to different datasets to identify users with high accuracy and low false alarm rates without any modification to the language model.     

Quantum‐Tunneling Metal‐Insulator‐Metal Diodes Made by Rapid Atmospheric Pressure Chemical Vapor Deposition

A quantum‐tunneling metal‐insulator‐metal (MIM) diode is fabricated by atmospheric pressure chemical vapor deposition (AP‐CVD) for the first time. This scalable method is used to produce MIM diodes with high‐quality, pinhole‐free Al₂O₃ films more rapidly than by conventional vacuum‐based approaches. This work demonstrates that clean room fabrication is not a prerequisite for quantum‐enabled devices. In fact, the MIM diodes fabricated by AP‐CVD show a lower effective barrier height (2.20 eV) at the electrode–insulator interface than those fabricated by conventional plasma‐enhanced atomic layer deposition (2.80 eV), resulting in a lower turn on voltage of 1.4 V, lower zero‐bias resistance, and better asymmetry of 107.     

Biomimetic Natural Killer Membrane Camouflaged Polymeric Nanoparticle for Targeted Bioimaging

In the present study, a biomimetic nanoconstruct (BNc) with a multimodal imaging system is engineered using tumor homing natural killer cell membrane (NKM), near‐infrared (NIR) fluorescent dye, and gadolinium (Gd) conjugate‐based magnetic resonance imaging contrast agent onto the surface of a polymeric nanoparticle. The engineered BNc is 110 ± 20 nm in size and showed successful retention of NKM proteins. The magnetic properties of the BNc are found to be tunable from 2.1 ± 0.17 to 5.3 ± 0.5 mm ^?1 s^?1 under 14.1 T, by adjusting the concentration of Gd‐lipid conjugate onto the surface of the BNc. Confocal imaging and cell sorting analysis reveal a distinguishable cellular interaction of the BNc with MCF‐7 cells in comparison to that of bare polymeric nanoparticles suggesting the tumor homing properties of NKM camouflage system. The in vitro cellular interaction results are further confirmed by in vivo NIR fluorescent tumor imaging and ex vivo MR imaging, respectively. Pharmacokinetics and biodistribution analysis of the BNc show longer circulation half‐life (≈9.5 h) and higher tumor accumulation (10% of injected dose) in MCF‐7 induced tumor‐bearing immunodeficient NU/NU nude mice. Owing to the proven immunosurveillance potential of NK‐cell in the field of immunotherapy, the BNc engineered herein would hold promises in the design consideration of nanomedicine engineering.