458 Hits in 1.7 sec

Curriculum Offline Imitation Learning [article]

Minghuan Liu, Hanye Zhao, Zhengyu Yang, Jian Shen, Weinan Zhang, Li Zhao, Tie-Yan Liu
2022 arXiv   pre-print
Offline reinforcement learning (RL) tasks require the agent to learn from a pre-collected dataset with no further interactions with the environment. Despite the potential to surpass the behavioral policies, RL-based methods are generally impractical due to the training instability and bootstrapping the extrapolation errors, which always require careful hyperparameter tuning via online evaluation. In contrast, offline imitation learning (IL) has no such issues since it learns the policy directly
more » ... without estimating the value function by bootstrapping. However, IL is usually limited in the capability of the behavioral policy and tends to learn a mediocre behavior from the dataset collected by the mixture of policies. In this paper, we aim to take advantage of IL but mitigate such a drawback. Observing that behavior cloning is able to imitate neighboring policies with less data, we propose Curriculum Offline Imitation Learning (COIL), which utilizes an experience picking strategy for imitating from adaptive neighboring policies with a higher return, and improves the current policy along curriculum stages. On continuous control benchmarks, we compare COIL against both imitation-based and RL-based methods, showing that it not only avoids just learning a mediocre behavior on mixed datasets but is also even competitive with state-of-the-art offline RL methods.
arXiv:2111.02056v2 fatcat:d7obnsol55bdlnec7gim5juzuq

Parallel Graph Connectivity in Log Diameter Rounds [article]

Alexandr Andoni, Clifford Stein, Zhao Song, Zhengyu Wang, Peilin Zhong
2018 arXiv   pre-print
We study graph connectivity problem in MPC model. On an undirected graph with n nodes and m edges, O( n) round connectivity algorithms have been known for over 35 years. However, no algorithms with better complexity bounds were known. In this work, we give fully scalable, faster algorithms for the connectivity problem, by parameterizing the time complexity as a function of the diameter of the graph. Our main result is a O( D _m/n n) time connectivity algorithm for diameter-D graphs, using Θ(m)
more » ... otal memory. If our algorithm can use more memory, it can terminate in fewer rounds, and there is no lower bound on the memory per processor. We extend our results to related graph problems such as spanning forest, finding a DFS sequence, exact/approximate minimum spanning forest, and bottleneck spanning forest. We also show that achieving similar bounds for reachability in directed graphs would imply faster boolean matrix multiplication algorithms. We introduce several new algorithmic ideas. We describe a general technique called double exponential speed problem size reduction which roughly means that if we can use total memory N to reduce a problem from size n to n/k, for k=(N/n)^Θ(1) in one phase, then we can solve the problem in O(_N/n n) phases. In order to achieve this fast reduction for graph connectivity, we use a multistep algorithm. One key step is a carefully constructed truncated broadcasting scheme where each node broadcasts neighbor sets to its neighbors in a way that limits the size of the resulting neighbor sets. Another key step is random leader contraction, where we choose a smaller set of leaders than many previous works do.
arXiv:1805.03055v1 fatcat:2epvlvyplrckjoiiiasigqzi44

Design of high reliability nuclear logging probe

Yingchun Xiao, Shihong Xiang, Zhengyu Zhao, Zuping Qian
2010 Procedia Engineering  
Nuclear Logging Probe may measure the naturally occurring gamma ray radiation level and detect gamma rays from a radioactive source within a well bore. It is used for depth correlation, lithology, mixture density measurement, identification of radioactive scale. But because of the harsh working environment, high temperature above 100 , high pressure, shock, it is difficult to get accurate measurement result. In this paper several possible cases that affect the measurement result were discussed;
more » ... and then provide appropriate solutions. The application of photomultiplier tube (PMT), NaI crystal, design of photon counting unit, divider of PMT were also discussed; A detailed circuit with high reliability is presented. Some measured results of the actual circuit are shown.
doi:10.1016/j.proeng.2010.11.035 fatcat:lkyzug7byfaw3dfpqx57v7aklm

(Nearly) Sample-Optimal Sparse Fourier Transform in Any Dimension; RIPless and Filterless [article]

Vasileios Nakos, Zhao Song, Zhengyu Wang
2019 arXiv   pre-print
[Sie] [Son17] Zhao Song. High dimensional Fourier transform in the continuous setting. In Manuscript, 2017. [Son19] Zhao Song. Matrix Theory : Optimization, Concentration and Algorithms.  ...  [CKPS16] Xue Chen, Daniel M Kane, Eric Price, and Zhao Song. Fourier-sparse interpolation without a frequency gap.  ... 
arXiv:1909.11123v1 fatcat:wixvodmifjb63f7ryh3cdm7xja

dualFace:Two-Stage Drawing Guidance for Freehand Portrait Sketching [article]

Zhengyu Huang, Yichen Peng, Tomohiro Hibino, Chunqi Zhao, Haoran Xie, Tsukasa Fukusato, Kazunori Miyata
2021 arXiv   pre-print
In this paper, we propose dualFace, a portrait drawing interface to assist users with different levels of drawing skills to complete recognizable and authentic face sketches. dualFace consists of two-stage drawing assistance to provide global and local visual guidance: global guidance, which helps users draw contour lines of portraits (i.e., geometric structure), and local guidance, which helps users draws details of facial parts (which conform to user-drawn contour lines), inspired by
more » ... al artist workflows in portrait drawing. In the stage of global guidance, the user draws several contour lines, and dualFace then searches several relevant images from an internal database and displays the suggested face contour lines over the background of the canvas. In the stage of local guidance, we synthesize detailed portrait images with a deep generative model from user-drawn contour lines, but use the synthesized results as detailed drawing guidance. We conducted a user study to verify the effectiveness of dualFace, and we confirmed that dualFace significantly helps achieve a detailed portrait sketch. see
arXiv:2104.12297v1 fatcat:4kcmhl5x65ghjam24fxiynqroq

A Newton Power Flow for Study Purpose

Yubin Yao, Benhua Gu, Hong Zhao, Zhengyu Xue
2012 Energy Procedia  
The Newton power flow program which employing sparse matrix techniques needs less computer storage and running time, but makes it harder by modifying the program to adapt to further study such as power system security and continuation power flow. To meet study purpose a less complicated but fast Newton power flow algorithm is presented. A great deal of unnecessary computation is exempted in the algorithm simply by adding some selection actions in the procedure of the Jacobian matrix formation
more » ... d that of Gauss elimination. The presented method needs a little more running time than the one using sparse matrix techniques, but it is much fast than the one without sparse matrix techniques and can satisfy the needs for study purpose. Experimental results validate the practicability of the presented method.
doi:10.1016/j.egypro.2012.01.218 fatcat:6az2jztfnfce5ongqsprvp5bdq

A high isolation switching unit for MRI system

Yingchun Xiao, Zhengyu Zhao, Zuping Qian, Hao Zhou
2010 Procedia Engineering  
The magnetic resonance imaging (MRI) system mainly consists of a large powerful magnet, a high power frequency tunable RF source, several receivers, L-C parallel resonance circuits, timing control unit, signal processing unit and a T/R switching unit. In order to protect the receivers from overload, and to obtain clear images the switching unit should meet the several requirements: low insertion, high isolation and fast switching speed. Such a switching unit with 1.5 Kilowatts handling
more » ... y is presented employing PIN diodes as key elements. This paper consists of four parts. 1) how to specify a PIN diode; 2) how to drive a PIN diode, and then presents a new method to drive PIN diode. It is no need to utilize neither optocoupler nor isolated power; 3) how to design a switch employing PIN diodes. 4) A simulation for the switch was performed to analyze its approximately specifications in advance and some measured results of the actual switch are shown.
doi:10.1016/j.proeng.2010.11.042 fatcat:o5xot5pyyfg6hp7dfhvwu5njem

Comprehensive Analysis of Codon Usage on Porcine Astrovirus

Huiguang Wu, Zhengyu Bao, Chunxiao Mou, Zhenhai Chen, Jingwen Zhao
2020 Viruses  
Porcine astrovirus (PAstV), associated with mild diarrhea and neurological disease, is transmitted in pig farms worldwide. The purpose of this study is to elucidate the main factors affecting codon usage to PAstVs. Phylogenetic analysis showed that the subtype PAstV-5 sat at the bottom of phylogenetic tree, followed by PAstV-3, PAstV-1, PAstV-2, and PAstV-4, indicating that the five existing subtypes (PAstV1-PAstV5) may be formed by multiple differentiations of PAstV ancestors. A codon usage
more » ... s was found in the PAstVs-2,3,4,5 from the analyses of effective number of codons (ENC) and relative synonymous codon usage (RSCU). Nucleotides A/U are more frequently used than nucleotides C/G in the genome CDSs of the PAstVs-3,4,5. Codon usage patterns of PAstV-5 are dominated by mutation pressure and natural selection, while natural selection is the main evolutionary force that affects the codon usage pattern of PAstVs-2,3,4. The analyses of codon adaptation index (CAI), relative codon deoptimization index (RCDI), and similarity index (SiD) showed the codon usage similarities between the PAstV and animals might contribute to the broad host range and the cross-species transmission of astrovirus. Our results provide insight into understanding the PAstV evolution and codon usage patterns.
doi:10.3390/v12090991 pmid:32899965 fatcat:n5yyzorwofb2pjikcvqxd5x74q

Unusual Enhancement of Midlatitude Sporadic-E Layers in Response to a Minor Geomagnetic Storm

Qiong Tang, Haiyang Sun, Zhitao Du, Jiaqi Zhao, Yi Liu, Zhengyu Zhao, Xueshang Feng
2022 Atmosphere  
This study investigates the variations of middle and low latitude sporadic-E (Es) layers in response to a geomagnetic storm. Es layers are observed by five ionosondes located in the Eastern Asian sector. The critical frequencies of Es layers (foEs) at six stations increased in sequence from high latitude stations to low latitude stations after IMF/Bz turning southward. Lomb–Scargle analysis shows the amplification of semidiurnal oscillation amplitude in the vertical height of Es layers during
more » ... omagnetic disturbance. Modeling results of the NCAR Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIEGCM) show the enhancement of the wind field in the mesosphere and the lower thermosphere (MLT) region. Our study provides evidence that the enhanced wind field in the MLT region during the storm period could result in the enhancement of Es layers at middle and low latitude.
doi:10.3390/atmos13050816 fatcat:nuirgt77grc7vmbniq46ntqidm

On the formation mechanism of Seasonal Persistence Barrier

Yishuai Jin, Zhengyu Liu, Chengfei He, Yuchu Zhao
2020 Journal of Climate  
Figure 7a suggests that variance peaks in the August and SPB occurs in June for North Pacific (Zhao et al. 2012; Liu et al. 2019) .  ...  Recent studies have also shown similar SPB features in many other areas such as the western North Pacific reemergence region (388-428N, 1608E-1808; Alexander et al. 1999 ) (Zhao et al. 2012) and Southern  ... 
doi:10.1175/jcli-d-19-0502.1 fatcat:mya5ggdp7zcinl4cgapz5dj6vy

Hybrid Online and Offline Reinforcement Learning for Tibetan Jiu Chess

Xiali Li, Zhengyu Lv, Licheng Wu, Yue Zhao, Xiaona Xu
2020 Complexity  
In this study, hybrid state-action-reward-state-action (SARSAλ) and Q-learning algorithms are applied to different stages of an upper confidence bound applied to tree search for Tibetan Jiu chess. Q-learning is also used to update all the nodes on the search path when each game ends. A learning strategy that uses SARSAλ and Q-learning algorithms combining domain knowledge for a feedback function for layout and battle stages is proposed. An improved deep neural network based on ResNet18 is used
more » ... or self-play training. Experimental results show that hybrid online and offline reinforcement learning with a deep neural network can improve the game program's learning efficiency and understanding ability for Tibetan Jiu chess.
doi:10.1155/2020/4708075 fatcat:nd4ncim3ybfsnantm4zfaslr3u

Human-Ocular-Physiological- Characteristics-Based Adaptive Console Design

Sheng Su, Sen Gu, Yiqian Zhao, Zhengyu Chen, Hanyu Wang, Wen Yang
2020 IEEE Access  
ZHENGYU CHEN is currently pursuing the Ph.D. degree in computer science with Zhejiang University. His research interests include meta learning and data mining.  ...  YIQIAN ZHAO received the B.S. degree in machine engineering from Jiangsu University, Zhenjiang, China, in 2010, the M.S. degree in industrial design engineering and the Ph.D. degree in machine engineering  ... 
doi:10.1109/access.2020.3002543 fatcat:lxhsnqmdtzfqzamowgfdonre7e

Playing for 3D Human Recovery [article]

Zhongang Cai, Mingyuan Zhang, Jiawei Ren, Chen Wei, Daxuan Ren, Jiatong Li, Zhengyu Lin, Haiyu Zhao, Shuai Yi, Lei Yang, Chen Change Loy, Ziwei Liu
2021 arXiv   pre-print
Image- and video-based 3D human recovery (i.e. pose and shape estimation) have achieved substantial progress. However, due to the prohibitive cost of motion capture, existing datasets are often limited in scale and diversity, which hinders the further development of more powerful models. In this work, we obtain massive human sequences as well as their 3D ground truths by playing video games. Specifically, we contribute, GTA-Human, a mega-scale and highly-diverse 3D human dataset generated with
more » ... he GTA-V game engine. With a rich set of subjects, actions, and scenarios, GTA-Human serves as both an effective training source. Notably, the "unreasonable effectiveness of data" phenomenon is validated in 3D human recovery using our game-playing data. A simple frame-based baseline trained on GTA-Human already outperforms more sophisticated methods by a large margin; for video-based methods, GTA-Human demonstrates superiority over even the in-domain training set. We extend our study to larger models to observe the same consistent improvements, and the study on supervision signals suggests the rich collection of SMPL annotations is key. Furthermore, equipped with the diverse annotations in GTA-Human, we systematically investigate the performance of various methods under a wide spectrum of real-world variations, e.g. camera angles, poses, and occlusions. We hope our work could pave way for scaling up 3D human recovery to the real world.
arXiv:2110.07588v1 fatcat:qvudr463cnb35m26tvrxijsg4e

Enantioselective Benzylation and Allylation of α-Trifluoromethoxy Indanones under Phase-Transfer Catalysis

Yumeng Liang, Mayaka Maeno, Zhengyu Zhao, Norio Shibata
2019 Molecules  
The organo-catalyzed enantioselective benzylation reaction of α-trifluoromethoxy indanones afforded α-benzyl-α-trifluoromethoxy indanones with a tetrasubstituted stereogenic carbon center in excellent yield with moderate enantioselectivity (up to 57% ee). Cinchona alkaloid-based chiral phase transfer catalysts were found to be effective for this transformation, and both enantiomers of α-benzyl-α-trifluoromethoxy indanones were accessed, depended on the use of cinchonidine and cinchonine-derived
more » ... catalyst. The method was extended to the enantioselective allylation reaction of α-trifluoromethoxy indanones to give the allylation products in moderate yield with good enantioselectivity (up to 76% ee).
doi:10.3390/molecules24152774 pmid:31366139 pmcid:PMC6696116 fatcat:klue6ed6obcwlhduzwga4tat7e

Quantum chemical predictions of vibrational mode analysis of germaformaldehyde and related molecules

Wenjuan Wang, Abraham F. Jalbout, Zhengyu Zhou, Shulei Zhao, Hongkun Ma
2007 Zenodo  
China E-mail : International Research Enterprises, 811 N Alvernon Rd., Tucson, Arizona 85711, USA and Department of Chemistry, University of Arizona, Tucson, Arizona 85721, USA  ... 
doi:10.5281/zenodo.5819617 fatcat:vlxhz2bfwzanli5qs5qsoxecgi
« Previous Showing results 1 — 15 out of 458 results