In mobile communication networks, host classification plays a critical role in constructing user profiles and ensuring network security. Traditional approaches, which rely on rule-based matching and shallow feature engineering, face significant limitations in coping with the high-frequency dynamic variations of hostnames and the labor-intensive maintenance of manual rules. To address these challenges, this paper proposes a novel frequency-aware hybrid-granularity tokenization method, specifically designed to capture both the semantic structure and statistical patterns of hostnames. By leveraging semi-supervised learning on large-scale host sequence data collected from real-world network environments, the proposed method enables effective service classification through vectorized host representations. This work not only offers an efficient and scalable solution for host analysis in personalized recommendation systems and mobile network security but also provides valuable insights into the design of pretrained tokenizers tailored for dynamic data scenarios.

Video Visual Relation Detection (VidVRD) is a pivotal task in the field of video analysis. It involves detecting object trajectories in videos, predicting potential dynamic relation between these trajectories, and ultimately representing these relationships in the form oftriplets. Correct prediction of relation is vital for VidVRD. Existing methods mostly adopt the simple fusion of visual and language features of entity trajectories as the feature representation for relation predicates. However, these methods do not take into account the dependency information between the relation predication and the subject and object within the triplet. To address this issue, we propose the entity dependency learning network(EDLN), which can capture the dependency information between relation predicates and subjects, objects, and subject-object pairs. It adaptively integrates these dependency information into the feature representation of relation predicates. Additionally, to effectively model the features of the relation existing between various object entities pairs, in the context encoding phase for relation predicate features, we introduce a fully convolutional encoding approach as a substitute for the self-attention mechanism in the Transformer. Extensive experiments on two public datasets demonstrate the effectiveness of the proposed EDLN.

Microcredit is a new financial instrument serving the segment of population that typically lack collateral and are highly likely to be rejected by traditional financial institutions, by lending very small loans. For platforms that participate in such consumer finance activities, the key challenge lies in risk management and the popular credit scoring method predicting whether a borrower would default or not takes an important role in this field. However, the fact is that we are often facing mass of raw data and the traditional credit scoring is heavily depending on feature engineering involving domain expert knowledge, intuition and trial and error, which is often time consuming. It is very challenging to derive effective features from raw data as the searching space can be very large with noninformative features. In this paper, we propose a new performance-driven framework automated generating discriminating features from raw data via reinforcement learning to help improve the default prediction of the downstream classifier which may be a logistic regression or boosting tree. Specially, we first define a formal paradigm for the automated feature derivation framework which unifies the feature structure, its interpretation and the calculation logic together. For the particularity of the financial industry, the interpretation of the feature is often of high interest. Then we reformulate the feature generation problem as reinforcement learning by constructing a transformation link and regarding it as a sequential decision process. In addition, we carry out an effective practice on default prediction in consumer finance. Finally, experimental results on the data of user behavior log from 360 Financial show the significant improvement of the proposed method over our years of domain expert knowledge and the Genetic Programming. Moreover, this FDRL framework can be easily adapted to other applications due to its versatility.

In this paper, a gain-scheduling, robust, and shared controller is proposed to assist drivers in tracking vehicle reference trajectory. In the controller, the driver steering parameters such as delay time, preview time, and steering gain are assumed to be varying with respect to the different characteristics of drivers, vehicle states, and driving scenarios. Meanwhile, the modeling errors and uncertainties in the tire cornering stiffness are also considered in the driver-vehicle system model and the controller design. A global objective function, considering the tracking error, the driver's physical and mental workloads, and the control effort, is designed to optimize the overall performance of the driver-vehicle system. Constraint on eigenvalue placement is added to the controller design to improve the performance of the closed-loop driver-vehicle system. Simulation results under different maneuvers show that the controller can significantly improve the system performance and reduce the driver's workloads. The controller can reduce the delay time of the driver-vehicle system in emergency maneuvers, particularly for inexperienced drivers.

ER fluid is a kind of functional fluid. The ER fluids have a unique characteristic, which changes its rheology characteristic to apply the electric fields. The property of the actuator using ER fluid is different from conventional actuators such as DC or AC servomotors. In this study, we developed a new actuator using particle-type ER fluid, which is called ER actuator. The developed actuator can produce two directional torques by using two ER clutches. We derive the simple models of ER actuator from reference torque to output torque and velocity characteristics. Then, we verified the validity of the models and control performance of the actuator by conducting velocity and position control experiments.

In industrial robots, each joint is driven through a reduction gear unit such as a harmonic drive gear. The elasticity of the drive systems cannot be ignored when the high-speed and precise motion control is required for robots. Due to the elasticity of the drive systems, vibratory behaviors are caused, and the control performances of robot systems deteriorate. In this study, we develop a new actuator that consists of an AC servomotor, a harmonic drive gear and an electrorheological (ER) damper. The ER damper can change the damping coefficient by applying an electric field. Since the gap between electrodes of the ER damper is 80 /spl mu/m only the voltage of 160 V is needed for the electric field of 2 kV/mm. The ER damper consumes little electric power. The actuator is installed on a one-link robot arm. The mathematical model of the robot arm is derived, and the eigenvalues of the robot system are analyzed. Experiments were carried out, and the effectiveness of the actuator is confirmed.

Since a direct-drive (DD) motor system does not have the reduction gear, its structure is simple and has stiffness. In the DD motor system, the disturbance torque directly influences the control accuracy. The high servo stiffness is effective and important for reducing the influence of the disturbance torque. However, the high servo control often makes the system vibratory or unstable. Therefore, the servo gain must be lowered for assuring stability, and the influence of the disturbance becomes serious. In this study, the DD motor system using an electro-rheological (ER) fluid damper is developed. By applying the ER damper to the DD motor system, the gain margin of the control system is improved, so it is possible to realize the high servo control. Finally, it is experimentally shown that the proposed system is effective for high-speed and high-precision control.

A semiclosed-loop control which utilizes the signal of an encoder mounted on a servomotor is adopted in the control of industrial robots. In this method, the position of the end-effector is not controlled very accurately. Therefore, we aim to apply a closed-loop control which uses the positional signal measured directly at the end-effector. Because of the flexibility of the driving system, however, it is very difficult to control the arm based on this method. In this study, a robot arm is developed which has a variable damper using an electrorheological fluid (ER fluid). It is shown that the ER damper is very effective for precise position control based on a closed-loop control method.

A variable viscous damper using electrorheological (ER) fluid is developed to suppress torsional vibration of industrial robot. An integrated system consisting of drive motor, harmonic drive, arm linkage and ER damper was measured experimentally and found to be able to reduce the resonance/antiresonance phenomenon by increasing the viscosity of ER damper. A position control system, which is based on both the conventional semi-closed loop control and this new type ER damper, is analyzed and the relation between the ER damping coefficient and the bandwidth of the system is discussed using pole assignment technique. It is shown that the ER damper provides both vibration suppression and disturbance rejection.

This paper is basic study on vibration control, disturbance rejection and friction compensation in robots with flexible driving systems. Firstly, the system gain characteristic at the antiresonance frequency is introduced to evaluate the vibratory behavior of the control system. Secondly, the joint torque negative feedback which has a good effect on vibration suppression is discussed, while it is shown that the property of disturbance rejection is easily deteriorated when using high-gain joint torque feedback. Based on the assignment of pole-distribution, the relations between the vibration suppression and the disturbance rejection are analyzed. Thirdly, a feedforward compensation control based on a friction observer is proposed, it is shown that the tracking error is effectively decreased by this friction compensation. Lastly, a position control system considering the trade-off between vibration suppression and disturbance rejection is presented, and experimental results are also given.