Search

Nowadays, people have an increasing interest in fresh products such as new shoes and cosmetics. To this end, an E-commerce platform Taobao launched a fresh-item hub page on the recommender system, with which customers can freely and exclusively explore and purchase fresh items, namely, the New Tendency page. In this work, we make a first attempt to tackle the fresh-item recommendation task with two major challenges. First, a fresh-item recommendation scenario usually faces the challenge that the training data are highly deficient due to low page views. In this paper, we propose a deep interest-shifting network (DisNet), which transfers knowledge from a huge number of auxiliary data and then shifts user interests with contextual information. Furthermore, three interpretable interest-shifting operators are introduced. Second, since the items are fresh, many of them have never been exposed to users, leading to a severe cold-start problem. Though this problem can be alleviated by knowledge transfer, we further babysit these fully cold-start items by a relational meta-Id-embedding generator (RM-IdEG). Specifically, it trains the item id embeddings in a learning-to-learn manner and integrates relational information for better embedding performance. We conducted comprehensive experiments on both synthetic datasets as well as a real-world dataset. Both DisNet and RM-IdEG significantly outperform state-of-the-art approaches, respectively. Empirical results clearly verify the effectiveness of the proposed techniques, which are arguably promising and scalable in real-world applications.

In view of the fact that music courses are seldom offered in schools and lack of management, there are many problems in the management system of music and art. In order to improve the learning effect of students in music courses, this paper proposes to build a music art management curriculum system based on the in-depth learning model. This paper puts forward the requirements of in-depth learning in music and art curriculum indicators and puts forward the model framework of music and art learning. Through the combined application of music courses, first of all, the current music course model is compared with Ghostnet music course, Mobilenet music course, and traditional music course to verify that the current music course model has certain advantages in accuracy, precision, recall, and image recognition performance. This proves that the current music curriculum is more suitable for the construction of this model. Secondly, the proposed model has a high score through the comprehensive application of several indicators in the course. Finally, under the model constructed by deep learning, the opening of music courses in schools has also increased significantly and improves the interest point and the application of people on the music art course so as to achieve the final application effect.

In this demo, we present an online multi-modal community query engine (MQE) on Alibaba's billion-scale heterogeneous network. MQE has two distinct features in comparison with existing community query engines. Firstly, MQE supports multimodal community search on heterogeneous graphs with keyword and image queries. Secondly, to facilitate community understanding in real business scenarios, MQE generates natural language descriptions for the retrieved community in combination with other useful demographic information. The distinct features of MQE benefit many downstream applications in Alibaba's e-commerce platform like recommendation. Our experiments confirm the effectiveness and efficiency of MQE on graphs with billions of edges.

Recommendations based on sequential User behaviors have become more and more common. Traditional methods depend on the premise of Markov processes and consider user behavior sequences as interests. However, they usually ignore the mining and representation of implicit features. Recently, recurrent neural networks (RNNs) have been adopted to leverage their power in modeling sequences and consider the dynamics of user behaviors. In order to better locate user preference, we design a network featuring Attention with Long-term Interval-based Gated Recurrent Units (ALI-GRU) to model temporal sequences of user actions. In the network, we propose a time interval-based GRU architecture to better capture long-term preferences and short-term intents when encoding user actions rather than the original GRU. And a specially matrix-form attention function is designed to learn weights of both long-term preferences and short-term user intents automatically. Experimental results on two well-known public datasets show that the proposed ALI-GRU achieves significant improvement than state-of-the-art RNN-based methods.

Fraud sellers in e-commerce usually promote their products via fake transactions. Such behaviors damage the reputation of the e-commerce platform and jeopardize the business environment in the platform. The search engine of existing e-commerce platforms mainly focuses on generating transactions by matching users' queries and sellers' products. The most common method to defense fraud sellers is to set up a blacklist based on fraud detection and manual investigation, and then punish those sellers in the list, which is inefficient and can only cover a small fraction of potential fraud sellers. In this paper, we propose the first fraud aware impression regulation system (FAIR) which is data-driven and can work in large-scale e-commerce platforms. Its main function is to actively regulate the impressions received by all potential fraud sellers in a real-time fashion. It utilizes the reinforcement learning architecture to dynamically adjust the impression regulation strategy under different reward settings, which can not only promote the impression regulation effects, but also improve the revenue of the platform simultaneously. We deploy FAIR on the Taobao platform of Alibaba, one of the world's largest e-commerce search platform, and perform an A/B test for two weeks. The results show that FAIR can effectively reduce the fraud impressions and improve the overall platform revenue at the same time.

This paper proposed a novel slot-structure z-axis gyroscope with a positive or negative electrostatic spring. The slot-structure gyroscope has in-plane drive and sense modes. Different from conventional comb finger structure, variable-area fixed out-of-plane electrodes under the proof-mass are used to excite the in-plane vibratory modes. To overcome the disadvantage that the resonant frequencies are unable to be adjusted through variable-area capacitors in the conventional slot-structure gyroscope, a variable-area capacitor with triangular shape fixed electrodes is designed to introduce the electrostatic tuning of the in-plane vibratory modes. A voltage difference Vt across the triangular fixed electrodes and movable electrodes produces a positive or negative electrostatic spring. The induced electrostatic spring constant is proportional to Vt2. A prototype with a positive electrostatic spring is fabricated and tested. The difference of the nature frequencies of the drive mode and sense mode is about 13.5 Hz. A 15 V Vt makes the gyroscope perfect mode matching. The scale factor of the prototype is about 72 mV/°/s when Vt is 0 V and increases with Vt. It presents a scale factor of 320.6 mV/°/s and noise density of 11.63 °/h/vHz when mode matched. Interestingly, the bandwidth of the gyroscope only slightly decreased from the initial value of 7 Hz to 4.6 Hz when mode matched, possible reasons are discussed in this paper.

This paper proposes a new oscillation control algorithm for vibratory micro-machined gyroscope system. The control algorithm tunes the resonant frequency of the gyroscope to the specified resonant frequency and maintains the specified amplitude of oscillation by altering the gyroscope dynamics. Feedback control is designed to achieve frequency tuning in the presence of time-delay. The stability of the control system is analyzed using the averaging method. Based on the real system, simulation models were set up to evaluate the control algorithm. Further, the algorithm mainly implemented using a field-programmable-gate-array (FPGA) is tested on a capacitive vibrational gyroscope operating at atmospheric pressure. First measurement results of the novel digital prototype show the phase variance of the detected signal of drive axis is 13 ppm in an hour while the amplitude variance is 22 ppm. A bias instability value of 6°/h is achieved using this new oscillation control loop.

We analyze the effects of possible noise sources on a fence structure micromachined capacitive accelerometer system by modeling and simulation to improve its performance. Simulation results show that a mismatch between the two initial sensing capacitors of the accelerometer or a mismatch between the two capacitance-voltage conversion circuits has a great effect on the output noise floor. When there is a serious mismatch, the noise induced by a sinusoidal carrier is the major noise source. When there is no or only a slight mismatch, the differential capacitance-voltage conversion circuits become the main noise source. The simulation results were validated by experiments and some effective approaches are proposed to improve the system resolution.

An introduction of a bulk micromachined frame gyroscope and the efforts made to lower the footing effect during deep reactive ion etching (DRIE) is featured. The gyroscope is driven into oscillation electrostatically and senses rate signal by the varying overlapped area of the electrodes, which eliminates the nonlinear behaviour of sensing the varying gap. Damping factors for the driving and sensing modes are low due to the slide-film damping, resulting in high quality factors in the atmosphere. A large proofmass can be made by DRIE, which also ensures a large pull-in voltage. The footing effect during fabrication has been greatly reduced by the improved process flow, which also increases the proof mass by 50%. Thus device quality and yield rate are both improved. Functional results of an improved prototype are given.

Variable-area capacitor with tuning capability is used to control the drive mode of MEMS gyroscopes for constant-frequency oscillation. The proposed control approach tunes the drive axis natural frequency to a preselected frequency and regulates the amplitude of the drive axis vibration. The algorithm mainly implemented using a field-programmable-gate-array (FPGA) is tested on a MEMS gyroscope operating at atmospheric pressure which is of our own manufacture. First measurement results of the novel digital prototype show the amplitude variance of the detected signal of drive axis is 26ppm in an hour while the phase variance is 45ppm.

The paper introduces specific ways of optimizing Proportional-Integral-Derivative (PID) type closed-loop Microelectromechanical System (MEMS) accelerometer system performance under fixed boundaries. Parameters tuning is applied to enhancing loop gain, under which configuration the tested Total Noise Equivalent Acceleration (TNEA) is around 3.2ug-√Hz @10Hz with bandwidth over 200Hz. This result is approaching this system's performance limit.

Location and route tracking is important for visitors and travelers, and it mainly depends on GPS information. However, GPS devices are not usually carried with by the travelers. Mobile phone with digital camera is the common standing item for people. We try to analyze the photos from mobile and compared to the known scenic, then predict the user location and accomplish the route tracking according the time and spatial information. In this paper, we choose our university as the scenic, and get good performance on daytime.

A bulk micromachined gyroscope fabricated with an improved process flow is measured systematically. It achieves good performance at atmosphere. The driving resonant frequency is 1663.5Hz with a quality factor 195, and the sensing resonant frequency is 1677Hz with a quality factor 200. The R2-nonlinearity is 0.008% within the range of ±43°/s, which is limited by the digital to analogue converter, with a scale factor of 42.17mV/°/s. The scale factor is reduced to 7.4mV/°/s to perform linear test within a larger range, and the R2-nonlinearity is 0.032% within ±200°/s. The bias stability is 19.55°/h and the measured noise floor is 0.0125°/s/√Hz at 1Hz.