arXiv subjects

Dong Li

Publications and source records attributed to Dong Li.

At least 487 records · Page 27Linked to original sources

arXiv2020

FLAME: A Self-Adaptive Auto-labeling System for Heterogeneous Mobile Processors

How to accurately and efficiently label data on a mobile device is critical for the success of training machine learning models on mobile devices. Auto-labeling data on mobile devices is challenging, because data is usually incrementally generated and there is possibility of having unknown labels. Furthermore, the rich hardware heterogeneity on mobile devices creates challenges on efficiently executing auto-labeling workloads. In this paper, we introduce Flame, an auto-labeling system that can label non-stationary data with unknown labels. Flame includes a runtime system that efficiently schedules and executes auto-labeling workloads on heterogeneous mobile processors. Evaluating Flame with eight datasets on a smartphone, we demonstrate that Flame enables auto-labeling with high labeling accuracy and high performance.

cs.LG↗

arXiv2020

Remarks on a nonlocal transport

We consider a one dimensional nonlocal transport equation and its natural multi-dimensional analogues. By using a new pointwise inequality for the Hilbert transform, we give a short proof of a nonlinear inequality first proved by Córdoba, Córdoba and Fontelos in 2005. We also prove several new weighted inequalities for the Hilbert transform and various nonlinear versions. Some of these results generalize to a related family of nonlocal models.

math.AP↗

arXiv2020

Demystifying the Performance of HPC Scientific Applications on NVM-based Memory Systems

The emergence of high-density byte-addressable non-volatile memory (NVM) is promising to accelerate data- and compute-intensive applications. Current NVM technologies have lower performance than DRAM and, thus, are often paired with DRAM in a heterogeneous main memory. Recently, byte-addressable NVM hardware becomes available. This work provides a timely evaluation of representative HPC applications from the "Seven Dwarfs" on NVM-based main memory. Our results quantify the effectiveness of DRAM-cached-NVM for accelerating HPC applications and enabling large problems beyond the DRAM capacity. On uncached-NVM, HPC applications exhibit three tiers of performance sensitivity, i.e., insensitive, scaled, and bottlenecked. We identify write throttling and concurrency control as the priorities in optimizing applications. We highlight that concurrency change may have a diverging effect on read and write accesses in applications. Based on these findings, we explore two optimization approaches. First, we provide a prediction model that uses datasets from a small set of configurations to estimate performance at various concurrency and data sizes to avoid exhaustive search in the configuration space. Second, we demonstrate that write-aware data placement on uncached-NVM could achieve $2$x performance improvement with a 60% reduction in DRAM usage.

cs.DC↗

arXiv2020

Neighborhood Cognition Consistent Multi-Agent Reinforcement Learning

Social psychology and real experiences show that cognitive consistency plays an important role to keep human society in order: if people have a more consistent cognition about their environments, they are more likely to achieve better cooperation. Meanwhile, only cognitive consistency within a neighborhood matters because humans only interact directly with their neighbors. Inspired by these observations, we take the first step to introduce \emph{neighborhood cognitive consistency} (NCC) into multi-agent reinforcement learning (MARL). Our NCC design is quite general and can be easily combined with existing MARL methods. As examples, we propose neighborhood cognition consistent deep Q-learning and Actor-Critic to facilitate large-scale multi-agent cooperations. Extensive experiments on several challenging tasks (i.e., packet routing, wifi configuration, and Google football player control) justify the superior performance of our methods compared with state-of-the-art MARL approaches.

cs.AI↗

arXiv2019

Giant enhancement of critical current density at high field in superconducting (Li,Fe)OHFeSe films by Mn doping

Critical current density (Jc) is one of the major limiting factors for high field applications of iron-based superconductors. Here, we report that Mn-ions are successfully incorporated into nontoxic superconducting (Li,Fe)OHFeSe films. Remarkably, the Jc is significantly enhanced from 0.03 to 0.32 MA/cm^2 under 33 T, and the vortex pinning force density monotonically increases up to 106 GN/m^3, which is the highest record so far among all iron-based superconductors. Our results demonstrate that Mn incorporation is an effective method to optimize the performance of (Li,Fe)OHFeSe films, offering a promising candidate for high-field applications.

cond-mat.supr-con↗

arXiv2019

A Compact Source for Quasi-Periodic Pulsation in an M-class Solar Flare

Quasi-periodic pulsations (QPP) are usually found in the light curves of solar and stellar flares, they carry the features of time characteristics and plasma emission of the flaring core, and could be used to diagnose the coronas of the Sun and remote stars. In this study, we combined the Atmospheric Imaging Assembly (AIA) onboard the Solar Dynamics Observatory and the Nobeyama Radioheliograph (NoRH) to observe an M7.7 class flare occurred at active region 11520 on 19 July 2012. A QPP was detected both in the AIA $131\unit{Å}$ bandpass and the NoRH $17\unit{GHz}$ channel, it had a period of about four minutes. In the spatial distribution of Fourier power, we found that this QPP originated from a compact source and that it overlapped with the X-ray source above the loop top. The plasma emission intensities in the AIA $131\unit{Å}$ bandpass were highly correlated within this region. The source region is further segmented into stripes that oscillated with distinctive phases. Evidence in this event suggests that this QPP was likely to be generated by intermittent energy injection into the reconnection region.

astro-ph.SR↗

arXiv2019

A Concise Survey of G4U

The general four-component model-based decomposition with unitary transformation of coherency matrix (G4U) is a state-of-the-art four-component decomposition, which has received extensive attentions recently. A literature survey is carried out to indicate the overall influence, improvement, development, evaluation, and application of G4U. Totally, 137 literatures are found mentioning G4U in Google Scholar\c{opyright} until October 7, 2019, which can be attributed into 4 categories in terms of the degree of concentration and 17 subcategories according to the focus of attention. Among these literatures, 61 of them simply mention G4U mainly because it is a new four-component model-based decomposition, a typical model-based decomposition, or even a target decomposition. There are also 9 literatures which improve G4U and develop G4U-like decompositions with unitary transformation of coherency matrix. 20 literatures generally use G4U as a typical target decomposition for comparison or a pseudo-color visualization technique to show the performance of some developed approaches. There are also 47 literatures dedicated to critically evaluate and deeply apply G4U in the remote sensing of forestry, agriculture, wetland, snow, glaciated terrain, earth surface, manmade target, environment, and damages caused by earthquake, tsunami, and landside, which indicates the value and significance of G4U in the true sense.

eess.SP↗

arXiv2019

Spectroscopic and Stereoscopic Observations of the Solar Jets

We present a comprehensive study of a series of recurrent jets that occurred at the periphery of the NOAA active region 12114 on 2014 July 7. These jets were found to share the same source region and exhibited rotational motions as they propagated outward. The multi-wavelength imaging observations made by the AIA and {\it IRIS} telescopes reveal that some of the jets contain cool plasma only, while some others contain not only cool but also hot plasma. The Doppler velocities calculated from the {\it IRIS} spectra show a continuous evolution from blue to red shifts as the jet motions change from upward to downward. Additionally, some jets exhibit opposite Doppler shifts on their both sides, indicative of rotating motions along their axes. The inclination angle and three-dimensional velocity of the largest jet were inferred from the imaging and spectroscopic observations, which show a high consistence with those derived from the stereoscopic analysis using dual-perspective observations by {\it SDO}/AIA and {\it STEREO}-B/EUVI. By relating the jets to the local UV/EUV and full-disk {\it GOES} X-ray emission enhancements, we found that the previous five small-scale jets were triggered by five bright points while the last/largest one was triggered by a C1.6 solar flare. Together with a number of type III radio bursts generated during the jet eruptions as well as a weak CME that was observed in association with the last jet, our observations provide evidences in support of multi-scale magnetic reconnection processes being responsible for the production of jet events.

astro-ph.SR↗

arXiv2019

UMap: Enabling Application-driven Optimizations for Page Management

Leadership supercomputers feature a diversity of storage, from node-local persistent memory and NVMe SSDs to network-interconnected flash memory and HDD. Memory mapping files on different tiers of storage provides a uniform interface in applications. However, system-wide services like mmap are optimized for generality and lack flexibility for enabling application-specific optimizations. In this work, we present Umap to enable user-space page management that can be easily adapted to access patterns in applications and storage characteristics. Umap uses the userfaultfd mechanism to handle page faults in multi-threaded applications efficiently. By providing a data object abstraction layer, Umap is extensible to support various backing stores. The design of Umap supports dynamic load balancing and I/O decoupling for scalable performance. Umap also uses application hints to improve the selection of caching, prefetching, and eviction policies. We evaluate Umap in five benchmarks and real applications on two systems. Our results show that leveraging application knowledge for page management could substantially improve performance. On average, Umap achieved 1.25 to 2.5 times improvement using the adapted configurations compared to the system service.

cs.DC↗

arXiv2019

Sentinel: Runtime Data Management on Heterogeneous Main MemorySystems for Deep Learning

Software-managed heterogeneous memory (HM) provides a promising solution to increase memory capacity and cost efficiency. However, to release the performance potential of HM, we face a problem of data management. Given an application with various execution phases and each with possibly distinct working sets, we must move data between memory components of HM to optimize performance. The deep neural network (DNN), as a common workload on data centers, imposes great challenges on data management on HM. This workload often employs a task dataflow execution model, and is featured with a large amount of small data objects and fine-grained operations (tasks). This execution model imposes challenges on memory profiling and efficient data migration. We present Sentinel, a runtime system that automatically optimizes data migration (i.e., data management) on HM to achieve performance similar to that on the fast memory-only system with a much smaller capacity of fast memory. To achieve this,Sentinel exploits domain knowledge about deep learning to adopt a custom approach for data management. Sentinel leverages workload repeatability to break the dilemma between profiling accuracy and overhead; It enables profiling and data migration at the granularity of data objects (not pages), by controlling memory allocation. This method bridges the semantic gap between operating system and applications. By associating data objects with the DNN topology, Sentinel avoids unnecessary data movement and proactively triggers data movement. Using only 20% of peak memory consumption of DNN models as fast memory size, Sentinel achieves the same or comparable performance (at most 8% performance difference) to that of the fast memory-only system on common DNN models; Sentinel also consistently outperforms a state-of-the-art solution by 18%.

cs.PF↗

arXiv2019

Artificial intelligence empowered multi-AGVs in manufacturing systems

AGVs are driverless robotic vehicles that picks up and delivers materials. How to improve the efficiency while preventing deadlocks is the core issue in designing AGV systems. In this paper, we propose an approach to tackle this problem.The proposed approach includes a traditional AGV scheduling algorithm, which aims at solving deadlock problems, and an artificial neural network based component, which predict future tasks of the AGV system, and make decisions on whether to send an AGV to the predicted starting location of the upcoming task,so as to save the time of waiting for an AGV to go to there first when the upcoming task is created. Simulation results show that the proposed method significantly improves the efficiency as against traditional method, up to 20% to 30%.

cs.AI↗

arXiv2019

Performance Analysis and Characterization of Training Deep Learning Models on Mobile Devices

Training deep learning models on mobile devices recently becomes possible, because of increasing computation power on mobile hardware and the advantages of enabling high user experiences. Most of the existing work on machine learning at mobile devices is focused on the inference of deep learning models (particularly convolutional neural network and recurrent neural network), but not training. The performance characterization of training deep learning models on mobile devices is largely unexplored, although understanding the performance characterization is critical for designing and implementing deep learning models on mobile devices. In this paper, we perform a variety of experiments on a representative mobile device (the NVIDIA TX2) to study the performance of training deep learning models. We introduce a benchmark suite and tools to study performance of training deep learning models on mobile devices, from the perspectives of memory consumption, hardware utilization, and power consumption. The tools can correlate performance results with fine-grained operations in deep learning models, providing capabilities to capture performance variance and problems at a fine granularity. We reveal interesting performance problems and opportunities, including under-utilization of heterogeneous hardware, large energy consumption of the memory, and high predictability of workload characterization. Based on the performance analysis, we suggest interesting research directions.

cs.LG↗

arXiv2019

Spam Review Detection with Graph Convolutional Networks

Customers make a lot of reviews on online shopping websites every day, e.g., Amazon and Taobao. Reviews affect the buying decisions of customers, meanwhile, attract lots of spammers aiming at misleading buyers. Xianyu, the largest second-hand goods app in China, suffering from spam reviews. The anti-spam system of Xianyu faces two major challenges: scalability of the data and adversarial actions taken by spammers. In this paper, we present our technical solutions to address these challenges. We propose a large-scale anti-spam method based on graph convolutional networks (GCN) for detecting spam advertisements at Xianyu, named GCN-based Anti-Spam (GAS) model. In this model, a heterogeneous graph and a homogeneous graph are integrated to capture the local context and global context of a comment. Offline experiments show that the proposed method is superior to our baseline model in which the information of reviews, features of users and items being reviewed are utilized. Furthermore, we deploy our system to process million-scale data daily at Xianyu. The online performance also demonstrates the effectiveness of the proposed method.

cs.IR↗

arXiv2019

MVB: A Large-Scale Dataset for Baggage Re-Identification and Merged Siamese Networks

In this paper, we present a novel dataset named MVB (Multi View Baggage) for baggage ReID task which has some essential differences from person ReID. The features of MVB are three-fold. First, MVB is the first publicly released large-scale dataset that contains 4519 baggage identities and 22660 annotated baggage images as well as its surface material labels. Second, all baggage images are captured by specially-designed multi-view camera system to handle pose variation and occlusion, in order to obtain the 3D information of baggage surface as complete as possible. Third, MVB has remarkable inter-class similarity and intra-class dissimilarity, considering the fact that baggage might have very similar appearance while the data is collected in two real airport environments, where imaging factors varies significantly from each other. Moreover, we proposed a merged Siamese network as baseline model and evaluated its performance. Experiments and case study are conducted on MVB.

cs.CV↗

arXiv2019

Modulating quantum fluctuations of scattered lights in disordered media via wavefront shaping

After multiple scattering of quadrature-squeezed lights in a disordered medium, the quadrature amplitudes of the scattered modes present an excess noise above the shot-noise level [Opt. Expr. 14, 6919 (2006)]. A natural question is raised whether there exists a method of suppressing the quadrature fluctuation of the output mode. The answer is affirmative. In this work, we prove that wavefront shaping is a promising method to reduce the quantum noise of quadrature amplitudes of the scattered modes. This reduction is owing to the destructive interference of quantum noise. Specifically, when the single-mode squeezed states are considered as inputs, the quantum fluctuation can always be reduced, even below the shot-noise level. These results may have potential applications in quantum information processing, for instance, sub-wavelength imaging using the scattering superlens with squeezed-state sources.

quant-ph↗

arXiv2019

Combining Q&A Pair Quality and Question Relevance Features on Community-based Question Retrieval

The Q&A community has become an important way for people to access knowledge and information from the Internet. However, the existing translation based on models does not consider the query specific semantics when assigning weights to query terms in question retrieval. So we improve the term weighting model based on the traditional topic translation model and further considering the quality characteristics of question and answer pairs, this paper proposes a communitybased question retrieval method that combines question and answer on quality and question relevance (T2LM+). We have also proposed a question retrieval method based on convolutional neural networks. The results show that Compared with the relatively advanced methods, the two methods proposed in this paper increase MAP by 4.91% and 6.31%.

cs.IR↗

arXiv2019

EasyCrash: Exploring Non-Volatility of Non-Volatile Memory for High Performance Computing Under Failures

Emerging non-volatile memory (NVM) is promising for building future HPC. Leveraging the non-volatility of NVM as main memory, we can restart the application using data objects remaining on NVM when the application crashes. This paper explores this solution to handle HPC under failures, based on the observation that many HPC applications have good enough intrinsic fault tolerance. To improve the possibility of successful recomputation with correct outcomes and ignorable performance loss, we introduce EasyCrash, a framework to decide how to selectively persist application data objects during application execution. Our evaluation shows that EasyCrash transforms 54% of crashes that cannot correctly recompute into the correct computation while incurring a negligible performance overhead (1.5% on average). Using EasyCrash and application intrinsic fault tolerance, 82% of crashes can successfully recompute. When EasyCrash is used with a traditional checkpoint scheme, it enables up to 24% improvement (15% on average) in system efficiency.

cs.PF↗

arXiv2019

Quantum coherence fraction

As an analogy of fully entangled fraction in the framework of entanglement theory, we have introduced the notion of quantum coherence fraction $C_{\mathcal{F}}$, which quantifies the closeness between a given state and the set of maximally coherent states. By providing an alternative formulation of the robustness of coherence $C_{\mathcal{R}}$, we have elucidated the relationship between quantum coherence fraction and the normalized version of $C_{\mathcal{R}}$ (i.e., $\overline{C}_{\mathcal{R}}$), where the role of genuinely incoherent operations (GIO) is highlighted. Numerical simulation shows that though as expected $C_{\mathcal{F}}$ is upper bounded by $\overline{C}_{\mathcal{R}}$, $C_{\mathcal{F}}$ constitutes a good approximation to $\overline{C}_{\mathcal{R}}$ especially in low-dimensional Hilbert spaces. Even more intriguingly, we can analytically prove that $C_{\mathcal{F}}$ is exactly equivalent to $\overline{C}_{\mathcal{R}}$ for qubit and qutrit states. Moreover, some intuitive properties and implications of $C_{\mathcal{F}}$ are also indicated.

quant-ph↗