arXiv subjects

Yang Yu

Publications and source records attributed to Yang Yu.

At least 361 records · Page 20Linked to original sources

arXiv2021

HieRec: Hierarchical User Interest Modeling for Personalized News Recommendation

User interest modeling is critical for personalized news recommendation. Existing news recommendation methods usually learn a single user embedding for each user from their previous behaviors to represent their overall interest. However, user interest is usually diverse and multi-grained, which is difficult to be accurately modeled by a single user embedding. In this paper, we propose a news recommendation method with hierarchical user interest modeling, named HieRec. Instead of a single user embedding, in our method each user is represented in a hierarchical interest tree to better capture their diverse and multi-grained interest in news. We use a three-level hierarchy to represent 1) overall user interest; 2) user interest in coarse-grained topics like sports; and 3) user interest in fine-grained topics like football. Moreover, we propose a hierarchical user interest matching framework to match candidate news with different levels of user interest for more accurate user interest targeting. Extensive experiments on two real-world datasets validate our method can effectively improve the performance of user modeling for personalized news recommendation.

cs.IR↗

arXiv2021

Sand Creep Motion in Slow Spin-up Experiment: An Analogue of Regolith Migration on Asteroids

We studied the creep motion of granular materials in a gradient potential field that is created using a slow spin-up experiment device. Natural sand confined in the acrylic box is spun up by a controlled turntable and the surface flows are captured using video-based measurements. Various spin-up accelerations were considered to understand the responses of creep motion on different accelerating paths. Convergent behaviors in the morphological change of sand surface were observed in the final steady state. To quantify the quasi-static spin-up process, we examined the net flux and the surface slope as a function of the spin rate and offset from the rotation axis. Evolution of sand creep motion demonstrated behaviors similar to regolith migration in numeric simulations, showing intermittency like general sheared granular systems. We noticed the sand surface approaches criticality as the spin-up proceeding, consistent with the observation that top-shaped asteroids near limiting spin rate take on critical shape. Comparisons to large-scale numeric simulations and analytical solutions reveal underlying similarities between our experiments and the million-year evolution of asteroid regolith under YORP acceleration, which raises the possibility of studying asteroid surface processes in laboratory analogue experiments.

astro-ph.EP↗

arXiv2021

Learning Action-Transferable Policy with Action Embedding

Transfer learning (TL) is a promising way to improve the sample efficiency of reinforcement learning. However, how to efficiently transfer knowledge across tasks with different state-action spaces is investigated at an early stage. Most previous studies only addressed the inconsistency across different state spaces by learning a common feature space, without considering that similar actions in different action spaces of related tasks share similar semantics. In this paper, we propose a method to learning action embeddings by leveraging this idea, and a framework that learns both state embeddings and action embeddings to transfer policy across tasks with different state and action spaces. Our experimental results on various tasks show that the proposed method can not only learn informative action embeddings but accelerate policy learning.

cs.LG↗

arXiv2021

Exact Higher-order Bulk-boundary Correspondence of Corner-localized States

We demonstrate that the presence of a localized state at the corner of an insulating domain is not always a predictor of a certain non-trivial higher-order topological invariant, even though they appear to co-exist in the same Hamiltonian parameter space. Our analysis of $C^n$-symmetric crystalline insulators and their multi-layer stacks reveals that topological corner states are not necessarily correlated with other well-established higher-order boundary observables, such as fractional corner charge or filling anomaly. In a $C^3$-symmetric breathing Kagome lattice, for example, we show that the bulk polarization, which successfully predicts the fractional corner anomaly, fails to be the relevant topological invariant for zero-energy corner states; instead, these corner states can be exactly explained by the decoration of topological edges. Also, while the zero-energy corner states in $C^4$-symmetric topological crystalline insulators have long been conjectured to be the result of the bulk polarization at quarter-filling, we correct this misconception by introducing a proper bulk invariant at half-filling and establishing a precise bulk-corner correspondence. By refining several bulk-corner correspondences in two-dimensional topological crystalline insulators, our work motivates further development of rigorous theoretical grounds for associating the existence of corner states with higher-order topology of host materials.

cond-mat.mes-hall↗

arXiv2021

ASBSO: An Improved Brain Storm Optimization With Flexible Search Length and Memory-Based Selection

Brain storm optimization (BSO) is a newly proposed population-based optimization algorithm, which uses a logarithmic sigmoid transfer function to adjust its search range during the convergent process. However, this adjustment only varies with the current iteration number and lacks of flexibility and variety which makes a poor search effciency and robustness of BSO. To alleviate this problem, an adaptive step length structure together with a success memory selection strategy is proposed to be incorporated into BSO. This proposed method, adaptive step length based on memory selection BSO, namely ASBSO, applies multiple step lengths to modify the generation process of new solutions, thus supplying a flexible search according to corresponding problems and convergent periods. The novel memory mechanism, which is capable of evaluating and storing the degree of improvements of solutions, is used to determine the selection possibility of step lengths. A set of 57 benchmark functions are used to test ASBSO's search ability, and four real-world problems are adopted to show its application value. All these test results indicate the remarkable improvement in solution quality, scalability, and robustness of ASBSO.

cs.NE↗

arXiv2021

TurboTransformers: An Efficient GPU Serving System For Transformer Models

The transformer is the most critical algorithm innovation of the Nature Language Processing (NLP) field in recent years. Unlike the Recurrent Neural Network (RNN) models, Transformers can process on dimensions of sequence lengths in parallel, therefore leading to better accuracy on long sequences. However, efficient deployments of them for online services in data centers equipped with GPUs are not easy. First, more computation introduced by transformer structures makes it more challenging to meet the latency and throughput constraints of serving. Second, NLP tasks take in sentences of variable length. The variability of input dimensions brings a severe problem to efficient memory management and serving optimization. This paper designed a transformer serving system called TurboTransformers, which consists of a computing runtime and a serving framework to solve the above challenges. Three innovative features make it stand out from other similar works. An efficient parallel algorithm is proposed for GPU-based batch reduction operations, like Softmax and LayerNorm, major hot spots besides BLAS routines. A memory allocation algorithm, which better balances the memory footprint and allocation/free efficiency, is designed for variable-length input situations. A serving framework equipped with a new batch scheduler using dynamic programming achieves the optimal throughput on variable-length requests. The system can achieve the state-of-the-art transformer model serving performance on GPU platforms and can be seamlessly integrated into your PyTorch code with a few lines of code.

cs.DC↗

arXiv2021

The Role of the Hercules Autonomous Vehicle During the COVID-19 Pandemic: An Autonomous Logistic Vehicle for Contactless Goods Transportation

Since early 2020, the coronavirus disease 2019 (COVID-19) has spread rapidly across the world. As at the date of writing this article, the disease has been globally reported in 223 countries and regions, infected over 108 million people and caused over 2.4 million deaths (https://covid19.who.int/, accessed on Feb. 17, 2021). Avoiding person-to-person transmission is an effective approach to control and prevent the pandemic. However, many daily activities, such as transporting goods in our daily life, inevitably involve person-to-person contact. Using an autonomous logistic vehicle to achieve contact-less goods transportation could alleviate this issue. For example, it can reduce the risk of virus transmission between the driver and customers. Moreover, many countries have imposed tough lockdown measures to reduce the virus transmission (e.g., retail, catering) during the pandemic, which causes inconveniences for human daily life. Autonomous vehicle can deliver the goods bought by humans, so that humans can get the goods without going out. These demands motivate us to develop an autonomous vehicle, named as Hercules, for contact-less goods transportation during the COVID-19 pandemic. The vehicle is evaluated through real-world delivering tasks under various traffic conditions.

cs.RO↗

arXiv2021

Self-supervised learning for fast and scalable time series hyper-parameter tuning

Hyper-parameters of time series models play an important role in time series analysis. Slight differences in hyper-parameters might lead to very different forecast results for a given model, and therefore, selecting good hyper-parameter values is indispensable. Most of the existing generic hyper-parameter tuning methods, such as Grid Search, Random Search, Bayesian Optimal Search, are based on one key component - search, and thus they are computationally expensive and cannot be applied to fast and scalable time-series hyper-parameter tuning (HPT). We propose a self-supervised learning framework for HPT (SSL-HPT), which uses time series features as inputs and produces optimal hyper-parameters. SSL-HPT algorithm is 6-20x faster at getting hyper-parameters compared to other search based algorithms while producing comparable accurate forecasting results in various applications.

cs.LG↗

arXiv2021

Derivative-Free Reinforcement Learning: A Review

Reinforcement learning is about learning agent models that make the best sequential decisions in unknown environments. In an unknown environment, the agent needs to explore the environment while exploiting the collected information, which usually forms a sophisticated problem to solve. Derivative-free optimization, meanwhile, is capable of solving sophisticated problems. It commonly uses a sampling-and-updating framework to iteratively improve the solution, where exploration and exploitation are also needed to be well balanced. Therefore, derivative-free optimization deals with a similar core issue as reinforcement learning, and has been introduced in reinforcement learning approaches, under the names of learning classifier systems and neuroevolution/evolutionary reinforcement learning. Although such methods have been developed for decades, recently, derivative-free reinforcement learning exhibits attracting increasing attention. However, recent survey on this topic is still lacking. In this article, we summarize methods of derivative-free reinforcement learning to date, and organize the methods in aspects including parameter updating, model selection, exploration, and parallel/distributed methods. Moreover, we discuss some current limitations and possible future directions, hoping that this article could bring more attentions to this topic and serve as a catalyst for developing novel and efficient approaches.

cs.LG↗

arXiv2021

NeoRL: A Near Real-World Benchmark for Offline Reinforcement Learning

Offline reinforcement learning (RL) aims at learning a good policy from a batch of collected data, without extra interactions with the environment during training. However, current offline RL benchmarks commonly have a large reality gap, because they involve large datasets collected by highly exploratory policies, and the trained policy is directly evaluated in the environment. In real-world situations, running a highly exploratory policy is prohibited to ensure system safety, the data is commonly very limited, and a trained policy should be well validated before deployment. In this paper, we present a near real-world offline RL benchmark, named NeoRL, which contains datasets from various domains with controlled sizes, and extra test datasets for policy validation. We evaluate existing offline RL algorithms on NeoRL and argue that the performance of a policy should also be compared with the deterministic version of the behavior policy, instead of the dataset reward. The empirical results demonstrate that the tested offline RL algorithms become less competitive to the deterministic policy on many datasets, and the offline policy evaluation hardly helps. The NeoRL suit can be found at http://polixir.ai/research/neorl. We hope this work will shed some light on future research and draw more attention when deploying RL in real-world systems.

cs.LG↗

arXiv2021

Active millimeter wave three-dimensional scan real-time imaging mechanism with a line antenna array

Active Millimeter wave (AMMW) imaging is of interest as it has played important roles in wide variety of applications, from nondestructive test to medical diagnosis. Current AMMW imaging systems have a high spatial resolution and can realize three-dimensional (3D) imaging. However, conventional AMMW imaging systems based on the synthetic aperture require either time-consume acquisition or reconstruction. The AMMW imaging systems based on real-aperture are able to real-time imaging but they need a large aperture and a complex two-dimensional (2D) scan structure to get 3D images. Besides, most AMMW imaging systems need the targets keep still and hold a special posture while screening, limiting the throughput. Here, by using beam control techniques and fast post-processing algorithms, we demonstrate the AMMW 3D scan real-time imaging mechanism with a line antenna array, which can realize 3D real-time imaging by a simple one-dimensional (1D) linear moving, simultaneously, with a satisfactory throughput (over 2000 people per-hour, 10 times than the commercial AMMW imaging systems) and a low system cost. First, the original spherical beam lines generated by the linear antenna array are modulated to fan beam lines via a bi-convex cylindrical lens. Then the holographic imaging algorithm is used to primarily focus the echo data of the imaged object. Finally, the defocus blur is corrected rapidly to get high resolution images by deconvolution. Since our method does not need targets to keep still, has a low system cost, can achieve 3D real-time imaging with a satisfactory throughput simultaneously, this work has the potential to serve as a foundation for future short-range AMMW imaging systems, which can be used in a variety of fields such as security inspection, medical diagnosis, etc.

eess.IV↗

arXiv2021

Exponential convergence of distributed optimization for heterogeneous linear multi-agent systems

In this work we study a distributed optimal output consensus problem for heterogeneous linear multi-agent systems where the agents aim to reach consensus with the purpose of minimizing the sum of private convex costs. Based on output feedback, a fully distributed control law is proposed by using the proportional-integral (PI) control technique. For strongly convex cost functions with Lipschitz gradients, the designed controller can achieve convergence exponentially in an undirected and connected network. Furthermore, to remove the requirement of continuous communications, the proposed control law is then extended to periodic and event-triggered communication schemes, which also achieve convergence exponentially. Two simulation examples are given to verify the proposed control algorithms.

eess.SY↗

arXiv2020

AliExpress Learning-To-Rank: Maximizing Online Model Performance without Going Online

Learning-to-rank (LTR) has become a key technology in E-commerce applications. Most existing LTR approaches follow a supervised learning paradigm from offline labeled data collected from the online system. However, it has been noticed that previous LTR models can have a good validation performance over offline validation data but have a poor online performance, and vice versa, which implies a possible large inconsistency between the offline and online evaluation. We investigate and confirm in this paper that such inconsistency exists and can have a significant impact on AliExpress Search. Reasons for the inconsistency include the ignorance of item context during the learning, and the offline data set is insufficient for learning the context. Therefore, this paper proposes an evaluator-generator framework for LTR with item context. The framework consists of an evaluator that generalizes to evaluate recommendations involving the context, and a generator that maximizes the evaluator score by reinforcement learning, and a discriminator that ensures the generalization of the evaluator. Extensive experiments in simulation environments and AliExpress Search online system show that, firstly, the classic data-based metrics on the offline dataset can show significant inconsistency with online performance, and can even be misleading. Secondly, the proposed evaluator score is significantly more consistent with the online performance than common ranking metrics. Finally, as the consequence, our method achieves a significant improvement (\textgreater$2\%$) in terms of Conversion Rate (CR) over the industrial-level fine-tuned model in online A/B tests.

cs.LG↗

arXiv2020

High-fidelity, high-scalability two-qubit gate scheme for superconducting qubits

High-quality two-qubit gate operations are crucial for scalable quantum information processing. Often, the gate fidelity is compromised when the system becomes more integrated. Therefore, a low-error-rate, easy-to-scale two-qubit gate scheme is highly desirable. Here, we experimentally demonstrate a new two-qubit gate scheme that exploits fixed-frequency qubits and a tunable coupler in a superconducting quantum circuit. The scheme requires less control lines, reduces crosstalk effect, simplifies calibration procedures, yet produces a controlled-Z gate in 30ns with a high fidelity of 99.5%, derived from the interleaved randomized benchmarking method. Error analysis shows that gate errors are mostly coherence limited. Our demonstration paves the way for large-scale implementation of high-fidelity quantum operations.

quant-ph↗

arXiv2020

Experimental Observation of Tensor Monopoles with a Superconducting Qudit

Monopoles play a center role in gauge theories and topological matter. There are two fundamental types of monopoles in physics: vector monopoles and tensor monopoles. Examples of vector monopoles include the Dirac monopole in 3D and Yang monopole in 5D, which have been extensively studied and observed in condensed matter or artificial systems. However, tensor monopoles are less studied, and their observation has not been reported. Here we experimentally construct a tunable spin-1 Hamiltonian to generate a tensor monopole and then measure its unique features with superconducting quantum circuits. The energy structure of a 4D Weyl-like Hamiltonian with three-fold degenerate points acting as tensor monopoles is imaged. Through quantum-metric measurements, we report the first experiment that measures the Dixmier-Douady invariant, the topological charge of the tensor monopole. Moreover, we observe topological phase transitions characterized by the topological Dixmier-Douady invariant, rather than the Chern numbers as used for conventional monopoles in odd-dimensional spaces.

quant-ph↗

arXiv2020

Interactive Search Based on Deep Reinforcement Learning

With the continuous development of machine learning technology, major e-commerce platforms have launched recommendation systems based on it to serve a large number of customers with different needs more efficiently. Compared with traditional supervised learning, reinforcement learning can better capture the user's state transition in the decision-making process, and consider a series of user actions, not just the static characteristics of the user at a certain moment. In theory, it will have a long-term perspective, producing a more effective recommendation. The special requirements of reinforcement learning for data make it need to rely on an offline virtual system for training. Our project mainly establishes a virtual user environment for offline training. At the same time, we tried to improve a reinforcement learning algorithm based on bi-clustering to expand the action space and recommended path space of the recommendation agent.

cs.LG↗

arXiv2020

Circles are like Ellipses, or Ellipses are like Circles? Measuring the Degree of Asymmetry of Static and Contextual Embeddings and the Implications to Representation Learning

Human judgments of word similarity have been a popular method of evaluating the quality of word embedding. But it fails to measure the geometry properties such as asymmetry. For example, it is more natural to say "Ellipses are like Circles" than "Circles are like Ellipses". Such asymmetry has been observed from a psychoanalysis test called word evocation experiment, where one word is used to recall another. Although useful, such experimental data have been significantly understudied for measuring embedding quality. In this paper, we use three well-known evocation datasets to gain insights into asymmetry encoding of embedding. We study both static embedding as well as contextual embedding, such as BERT. Evaluating asymmetry for BERT is generally hard due to the dynamic nature of embedding. Thus, we probe BERT's conditional probabilities (as a language model) using a large number of Wikipedia contexts to derive a theoretically justifiable Bayesian asymmetry score. The result shows that contextual embedding shows randomness than static embedding on similarity judgments while performing well on asymmetry judgment, which aligns with its strong performance on "extrinsic evaluations" such as text classification. The asymmetry judgment and the Bayesian approach provides a new perspective to evaluate contextual embedding on intrinsic evaluation, and its comparison to similarity evaluation concludes our work with a discussion on the current state and the future of representation learning.

cs.CL↗

arXiv2020

Application of Jordan Decomposition to Non-Hermitian Lattice Models with Spectrally-Isolated Lower Dimensional States

When analyzing non-Hermitian lattice systems, the standard eigenmode decomposition utilized for the analysis of Hermitian systems must be replaced by Jordan decomposition. This approach enables us to identify the correct number of the left and right eigenstates of a large finite-sized lattice system, and to form a complete basis for calculating the resonant excitation of the system. Specifically, we derive the procedure for applying Jordan decomposition to a system with spectrally-isolated states. We use a non-Hermitian quadrupole insulator with zero-energy corner states as an example of a large system whose dimensionality can be drastically reduced to derive a low-dimensional "defective" Hamiltonian describing such localized states. Counter-intuitive and non-local properties of the resonant response of the system near zero energy are explained using the Jordan decomposition approach. Depending on the excitation properties of the corner states, we classify our non-Hermitian quadrupolar insulator into three categories: trivial, near-Hermitian, and non-local.

cond-mat.mes-hall↗