Current Location:home > Browse



1. chinaXiv:202005.00043 [pdf]

DEED: A general quantization scheme for saving bits in communication

Tian Ye; Peijun Xiao; Ruoyu Sun
Subjects: Mathematics >> Control and Optimization.

Quantization is a popular technique to reduce communication in distributed optimization. Motivated by the classical work on inexact gradient descent (GD) \cite{bertsekas2000gradient}, we provide a general convergence analysis framework for inexact GD that is tailored for quantization schemes. We also propose a quantization scheme Double Encoding and Error Diminishing (DEED). DEED can achieve small communication complexity in three settings: frequent-communication large-memory, frequent-communication small-memory, and infrequent-communication (e.g. federated learning). More specifically, in the frequent-communication large-memory setting, DEED can be easily combined with Nesterov's method, so that the total number of bits required is $ \tilde{O}( \sqrt{\kappa} \log 1/\epsilon )$, where $\tilde{O}$ hides numerical constant and $\log \kappa $ factors. In the frequent-communication small-memory setting, DEED combined with SGD only requires $\tilde{O}( \kappa \log 1/\epsilon)$ number of bits in the interpolation regime. In the infrequent communication setting, DEED combined with Federated averaging requires a smaller total number of bits than Federated Averaging. All these algorithms converge at the same rate as their non-quantized versions, while using a smaller number of bits.

submitted time 2020-06-16 Hits7784Downloads745 Comment 0

2. chinaXiv:202004.00009 [pdf]

CAN Algorithm: An Individual Level Approach to identify Consequences and Norms Sensitivities and Overall Action/inaction Preferences in Moral Decision-making

Chuanjun Liu; Jiangqun Liao
Subjects: Psychology >> Psychological Measurement

Gawronski et al. (2017) developed a CNI model to measure an agent’s norms sensitivity, consequences sensitivity, and generalized inaction/action preferences when making moral decisions. However, the CNI model presupposed that an agent considers consequences—norms—generalized inaction/action preferences sequentially, which is untenable based on recent evidence. Moreover, the CNI model generates parameters at the group level based on binary categoric data. Hence, the C/N/I parameters cannot be used for correlation analyses or other conventional research designs. To solve these limitations, we developed the CAN algorithm to compute norms and consequences sensitivities and overall action/inaction preferences algebraically in a parallel manner. We re-analyzed the raw data of Gawronski et al.(2017) to test the methodological predictions. Our results demonstrate that: (1) the C parameter is approximately equal between the CNI model and CAN algorithm; (2) the N parameter under the CNI model approximately equals N/(1 – C) under the CAN algorithm; (3) the I parameter and A parameter are reversed around 0.5 – the larger the I parameter, the more the generalized inaction versus action preference and the larger the A parameter, the more overall action versus inaction preference; (4) tests of differences in parameters between groups with the CNI model and CAN algorithm led to almost the same statistical conclusion; (5) Parameters from the CAN algorithm can be used for correlational analyses and multiple comparisons, and this is an advantage over the parameters from the CNI model. The theoretical and methodological implications of our study were also discussed.

submitted time 2020-04-03 Hits6196Downloads604 Comment 0

3. chinaXiv:201910.00029 [pdf]

Stand up to Action: The Postural Effect on Deontological Responding and the Boundary Condition of Dual Process

刘传军; 廖江群
Subjects: Psychology >> Social Psychology

Present research aimed to uncover a postural effect on deontological responding and its boundary condition of dual process. Previous literatures demonstrated a possibility that people have a stronger controlled cognitive process propensity, thus being less deontological in standing than sitting postures, and that this postural effect can be moderated by dual process. We conducted two studies in which participant read dilemma scenarios and rated the morally acceptance and action intention of the utilitarian proposal after each scenario when sitting or standing. The hypothesized postural effect was verified in a field study (Study 1) and also replicated in an experimental study (Study 2). Compared with those in sitting postures, participants in standing postures approved more to the utilitarian proposal and became less deontological. Furthermore, the postural effect was dismissed when participants made moral decisions with a dual task to increase cognitive load and reversed when participants made moral decisions after deliberate consideration of the sacrificial proposal (Study 2). Thus, the postural effect was stable across field and experimental studies and moderated by dual process. The present research supports and extends the dual process morality theory by confirming that body posture can affect moral decision-making, and also offers a novel evidence confirming the moderating role of dual process on embodiment effects. It enriches our knowledge that morality is evolutionarily embodied in postures and dual process can moderate the embodiment effects.

submitted time 2020-01-14 Hits16856Downloads1209 Comment 0

4. chinaXiv:201905.00012 [pdf]

Transfer Learning for Scientific Data Chain Extraction in Small Chemical Corpus with BERT-CRF Model

Na Pang; Li Qian; Weimin Lyu; Jin-Dong Yang
Subjects: Computer Science >> Natural Language Understanding and Machine Translation

Abstract. Computational chemistry develops fast in recent years due to the rapid growth and breakthroughs in AI. Thanks for the progress in natural language processing, researchers can extract more fine-grained knowledge in publications to stimulate the development in computational chemistry. While the works and corpora in chemical entity extraction have been restricted in the biomedicine or life science field instead of the chemistry field, we build a new corpus in chemical bond field anno- tated for 7 types of entities: compound, solvent, method, bond, reaction, pKa and pKa value. This paper presents a novel BERT-CRF model to build scientific chemical data chains by extracting 7 chemical entities and relations from publications. And we propose a joint model to ex- tract the entities and relations simultaneously. Experimental results on our Chemical Special Corpus demonstrate that we achieve state-of-art and competitive NER performance.

submitted time 2019-05-12 Hits17916Downloads848 Comment 0

5. chinaXiv:201810.00102 [pdf]

Where is the Embodiment Effect? The Hierarchical Access Priority Model

Chuanjun Liu; Jiangqun Liao; Kaiping Peng
Subjects: Psychology >> Cognitive Psychology

There are detailed theories and abundant empirical results regarding embodied cognition. However, embodiment effects are undergoing a replication crisis. Based on the hierarchical structure of embodiment tasks and the dual process property of embodiment phenomena, we propose the hierarchical access priority model (HAP). According to HAP, the generation of embodiment effects depends on the access priority of embodied variables to unconscious processes, and embodiment effects from different hierarchy levels show a contravariant relationship between effect size and stability. Theoretically, the stability of an embodiment effect is partly determined by the hierarchy of the embodied variable, and dissociation of the dual process moderates the effect size. Empirically, the hierarchical linear model analytic method should be considered for embodied research; the embodied variable could be designed as a mediating or moderating variable, and other possible masked mediating variables should be considered. HAP offers an insightful theoretical perspective for the embodiment replication crisis.

submitted time 2019-03-25 Hits9622Downloads1587 Comment 0

6. chinaXiv:201810.00184 [pdf]

Simulating hydrological responses to climate change using dynamic and statistical downscaling methods: a case study in the Kaidu River Basin, Xinjiang, China

BA Wulong; DU Pengfei; LIU Tie; BAO Anming; LUO Min; Mujtaba HASSAN; QIN Chengxin
Subjects: Geosciences >> History of Geosciences

Climate change may affect water resources by altering various processes in natural ecosystems. Dynamic and statistical downscaling methods are commonly used to assess the impacts of climate change on water resources. Objectively, both methods have their own advantages and disadvantages. In the present study, we assessed the impacts of climate change on water resources during the future periods (2020–2029 and 2040–2049) in the upper reaches of the Kaidu River Basin, Xinjiang, China, and discussed the uncertainties in the research processes by integrating dynamic and statistical downscaling methods (regional climate models (RCMs) and general circulation modes (GCMs)) and utilizing these outputs. The reference period for this study is 1990–1999. The climate change trend is represented by three bias-corrected RCMs (i.e., Hadley Centre Global Environmental Model version 3 regional climate model (HadGEM3-RA), Regional Climate Model version 4 (RegCM4), and Seoul National University Meso-scale Model version 5 (SUN-MM5)) and an ensemble of GCMs on the basis of delta change method under two future scenarios (RCP4.5 and RCP8.5). We applied the hydrological SWAT (Soil and Water Assessment Tool) model which uses the RCMs/GCMs outputs as input to analyze the impacts of climate change on the stream flow and peak flow of the upper reaches of the Kaidu River Basin. The simulation of climate factors under future scenarios indicates that both temperature and precipitation in the study area will increase in the future compared with the reference period, with the largest increase of annual mean temperature and largest percentage increase of mean annual precipitation being of 2.4°C and 38.4%, respectively. Based on the results from bias correction of climate model outputs, we conclude that the accuracy of RCM (regional climate model) simulation is much better for temperature than for precipitation. The percentage increase in precipitation simulated by the three RCMs is generally higher than that simulated by the ensemble of GCMs. As for the changes in seasonal precipitation, RCMs exhibit a large percentage increase in seasonal precipitation in the wet season, while the ensemble of GCMs shows a large percentage increase in the dry season. Most of the hydrological simulations indicate that the total stream flow will decrease in the future due to the increase of evaporation, and the maximum percentage decrease can reach up to 22.3%. The possibility of peak flow increasing in the future is expected to higher than 99%. These results indicate that less water is likely to be available in the upper reaches of the Kaidu River Basin in the future, and that the temporal distribution of flow may become more concentrated.

submitted time 2018-10-29 From cooperative journals:《Journal of Arid Land》 Hits3757Downloads816 Comment 0

7. chinaXiv:201810.00002 [pdf]


Chao, Zhao; Can, Xiong; Man, Qiao; Tongwei, Yuan; Jing, Wang; Yunteng, Qu; XiaoQian, Wang; Fangyao, Zhou; Zhijun, Li; Qian, Xu; Shiqi, Wang; Min, Chen; Wenyu, Wang; Yafei, Li; Yuen, Wu; Yadong, Li
Subjects: Chemistry >> Inorganic Chemistry

在这里,我们报告了一种高温气体迁移策略来制备多相的单原子铁催化剂,并展示其优良的类天然酶的活性(定义为单原子酶,SAE)。 结果表明,Fe SAE的局部定义明确的FeN4位点与含血红素的天然酶的活性金属中心高度相似。 Fe SAE表现出超级高的类过氧化物酶,氧化酶和过氧化氢酶的活性,超过了Fe3O4纳米酶40倍。更重要的是,Fe SAE可以有效活化过氧化氢(H2O2)产生羟基自由基(?OH)。 这使得其可以作为一种多相化的类Fenton系统,可以应对不断增加的环境污染。 我们的研究结果开辟了一个新的人造材料家族-模仿天然

submitted time 2018-10-04 Hits2195Downloads953 Comment 0

8. chinaXiv:201803.00088 [pdf]

Attention-operated working memory representations determine visual selective attention

Ting Luo; Shimin Fu; Bochuan Mou; Zhencai Chen; Dandan Tang; Antao Chen
Subjects: Psychology >> Cognitive Psychology

It has been demonstrated that visual attention is guided by information actively maintained in working memory (WM). However, it remains unknown whether other operations (e.g. inhibition) on WM contents influence selective attention. This issue was investigated in a visual search task where WM contents with either operation (maintenance or inhibition) appeared as distractors in the search display. Behavioral results showed that search performance was slowed down for presenting the maintained contents, but speeded up for the inhibited contents. These results suggested an operation-directed selection of WM contents that visual attention was distinctively influenced by contents with different operation. These observations were further confirmed by the indexes of event-related potentials (ERPs). The inhibited WM contents were suppressed at sensory gating stage (i.e., suppressed P1 amplitude), while the maintained WM contents guided visual attention (i.e., enhanced N2pc amplitude). It seems that results from ERPs and behavior are integrated. The maintained contents guided visual attention that hindered performance of searching a target, while the inhibited contents screened attention that facilitated target searching in the other space. Besides, P3 component indexing updating of WM, which revealed comparable latency both for the maintained and the inhibited WM contents, but with longer latency than the neutral contents. These results suggested that the WM contents with distinct operations were expelled from the focus of executive attention after the onset of the search task, which might be necessary for the WM contents influencing selective attention. The current study reveals that operations of WM contents distinctively affect early selective attention to the matching contents, which sheds some light on the interaction between WM and visual attention.

submitted time 2018-03-09 Hits11741Downloads1026 Comment 0

9. chinaXiv:201801.00588 [pdf]

Inverse Problem and Variation Method to Optimize Cascade Heat Exchange Network in Central Heating System

ZHANG Yin1,2*, WEI Zhiyuan2, ZHANG Yinping2, WANG Xin2
Subjects: Dynamic and Electric Engineering >> Engineering Thermophysics

Urban heating in northern China accounts for 40% of total building energy usage. In central heating systems, heat is often transferred from heat source to users by the heat network where several heat exchangers are installed at heat source, substations and terminals respectively. For given overall heating capacity and heat source temperature, increasing the terminal fluid temperature is an effective way to improve the thermal performance of such cascade heat exchange network for energy saving. In this paper, the mathematical optimization model of the cascade heat exchange network with three-stage heat exchangers in series is established. Aim at maximizing the cold fluid temperature for given hot fluid temperature and overall heating capacity, the optimal heat exchange area distribution and the medium fluids' flow rates are determined through inverse problem and variation method. The preliminary results show that the heat exchange areas should be distributed equally for each heat exchanger. It also indicates that in order to improve the thermal performance of the whole system, more heat exchange areas should be allocated to the heat exchanger where flow rate difference between two fluids is relatively small. This work is important for guiding the optimization design of practical cascade heating systems.

submitted time 2018-01-24 From cooperative journals:《热科学学报》 Hits28578Downloads673 Comment 0

10. chinaXiv:201711.02257 [pdf]

Insight-HXMT observations of the first binary neutron star merger GW170817

TiPei Li; ShaoLin Xiong; ShuangNan Zhang; FangJun Lu; LiMing Song; XueLei Cao; Zhi Chang; Gang Chen; Li Chen; TianXiang Chen; Yong Chen; YiBao Chen; YuPeng Chen; Wei Cui; WeiWei Cui; JingKang Deng; YongWei Dong; YuanYuan Du; MinXue Fu; GuanHua Gao
Subjects: Physics >> General Physics: Statistical and Quantum Mechanics, Quantum Information, etc.

Finding the electromagnetic (EM) counterpart of binary compact star merger, especially the binary neutron star (BNS) merger, is critically important for gravitational wave (GW) astronomy, osmology and fundamental physics. On Aug. 17, 2017, Advanced LIGO and Fermi/GBM independently triggered the first BNS merger, GW170817, and its high energy EM counterpart, GRB 170817A, respectively, resulting in a global observation campaign covering gamma-ray, X-ray, UV, optical, IR, radio as well as neutrinos. The High Energy X-ray telescope (HE) onboard Insight-HXMT (Hard X-ray Modulation Telescope) is the unique high-energy gamma-ray telescope that monitored the entire GW localization area and especially the optical counterpart (SSS17a/AT2017gfo) with very large collection area (~1000 cm2) and microsecond time resolution in 0.2-5 MeV. In addition, Insight-HXMT quickly implemented a Target of Opportunity (ToO) observation to scan the GW localization area for potential X-ray emission from the GW source. Although Insight-HXMT did not detect any significant high energy (0.2-5 MeV) radiation from GW170817, its observation helped to confirm the nexpected weak and soft nature of GRB 170817A. Meanwhile, Insight-HXMT/HE provides one of the most stringent constraints (~10-7 to 10-6 erg/cm2/s) for both GRB170817A and any other possible precursor or extended emissions in 0.2-5 MeV, which help us to better understand the properties of EM radiation from this BNS merger. Therefore the observation of Insight-HXMT constitutes an important chapter in the full context of multi-wavelength and multi-messenger observation of this historical GW event.

submitted time 2017-11-10 Hits1147Downloads644 Comment 0

123  Last  Go  [3 Pages/ 28 Totals]