检索结果-内蒙古大学图书馆

Transfer with action embeddings for deep reinforcement learning

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Chen, Yu Chen, Yingfeng Hu, Zhipeng Yang, Tianpei Fan, Changjie Yu, Yang Hao, Jianye NetEase Fuxi AI Lab Zhejiang University China College of Intelligence and Computing Tianjin University China National Key Laboratory for Novel Software Technology Nanjing University China

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. Copyright © 2019, The Authors. All rights reserved.

关键词： Embeddings

Talking Head Generation for Media Interaction System with Feature Disentanglement

学校读者我要写书评

暂无评论

Talking Head Generation for Media Interaction System with Fe...

International Conference on Parallel and Distributed Systems (ICPADS)

作者： Lei Zhang Qingyang Chen Zhilei Liu College of Intelligence and Computing School of New Media and Communication Tianjin University Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Changchun P.R. China State Key Laboratory for Novel Software Technology Nanjing University P.R. China School of New Media and Communication Tianjin University Tianjin China College of Intelligence and Computing Tianjin University Tianjin China

The task of talking head generation for the media interaction system is to take images and audio clips of the target face as input, and generate a realistic video of the target synchronized with the audio. Most of the existing works directly take all the information in the image as input, which causes the problem of feature information redundancy. At the same time, in the stage of feature fusion, the image and audio information is directly spliced, and the relationship between the two modal information is ignored. To solve the problems, we initially implement the disentanglement of image by introducing a loss function to separate the image into identity features and content-related features. Besides, we introduce a multi-head selfattention mechanism to learn the relationship between the two modal information of image and audio, and implement the full fusion of multi-modal information. In addition, we validate the effectiveness of our model through extensive quantitative and qualitative analysis of two datasets. Extensive experiments show the superiority of the proposed model in all aspects.

关键词： Analytical models Redundancy Media Synchronization Task analysis Faces

Expander Code: A Scalab.e Erasure-Resilient Code to Keep up with Data Growth in Distributed Storage

学校读者我要写书评

暂无评论

Expander Code: A Scalable Erasure-Resilient Code to Keep up ...

IEEE International Performance computing and Communication Conference

作者： Guangping Xu Sheng Lin Hua Zhang Xing Guo Kai Shit Computer and Communication Engineering School Tianjin University of Technology Tianjin China 300384 Tianjin Key Lab. of Intelligence Computing and New Software Technology Tianjin China 300384

ISBN: (纸本)9781479932153

To ensure high reliability and storage efficiency, erasure codes are preferred in storage systems. With the prevalent of distributed storage systems such as clouds storage, how to design a scalab.e and efficient erasure-resilient code is challenging. We propose a scalab.e binary linear code to keep up with data growth which has the following properties. Given the group size k and the code block length n, the proposed code corrects any two bit erasures among the n bits. The redundancy overhead of the code is 2/(k + 2), and each data bit affects exactly 2 parity bits. As results of these properties, if a data bit is changed or added, only two parity bits need to be updated;and the recovery of an erasured bit requires accessing at most k other bits and the recovery of two erasured bits requires at most 2k other bits. We give the construction algorithm by the order expansion of regular graphs;moreover, we optimize the failure resilience during the construction procedure. Compared with existing codes, our proposed code has notable benefits in storage scalab.lity, redundancy overhead and I/O bandwidth. The deployment of the proposed code in distributed storage systems can be simple and practical.

关键词： Storage systems Erasure codes Scalab.lity Fault-tolerance Graph expansion

REINFORCED IN-CONTEXT BLACK-BOX OPTIMIZATION

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Song, Lei Gao, Chen-Xiao Xue, Ke Wu, Chenyang Li, Dong Hao, Jianye Zhang, Zongzhang Qian, Chao National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China Huawei Noah’s Ark Lab China School of Computing and Intelligence Tianjin University China

Black-Box Optimization (BBO) has found successful applications in many fields of science and engineering. Recently, there has been a growing interest in meta-learning particular components of BBO algorithms to speed up optimization and get rid of tedious hand-crafted heuristics. As an extension, learning the entire algorithm from data requires the least lab.r from experts and can provide the most flexibility. In this paper, we propose RIBBO, a method to reinforce-learn a BBO algorithm from offline data in an end-to-end fashion. RIBBO employs expressive sequence models to learn the optimization histories produced by multiple behavior algorithms and tasks, leveraging the in-context learning ability of large models to extract task information and make decisions accordingly. Central to our method is to augment the optimization histories with regret-to-go tokens, which are designed to represent the performance of an algorithm based on cumulative regret over the future part of the histories. The integration of regret-to-go tokens enables RIBBO to automatically generate sequences of query points that satisfy the user-desired regret, which is verified by its universally good empirical performance on diverse problems, including BBO benchmark functions, hyper-parameter optimization and robot control problems. Copyright © 2024, The Authors. All rights reserved.

关键词： Optimization algorithms

Prediction of anti-cancer drug synergy based on cross-matching network and cancer molecular subtypes

学校读者我要写书评

暂无评论

Computers in Biology and Medicine 2024年 175卷 108441-108441页

作者： Su, Ran Han, Jingyi Sun, Changming Zhang, Degan Geng, Jie Wang, Ping Zeng, Xiaoyan School of Computer Software College of Intelligence and Computing Tianjin University Tianjin China CSIRO Data61 EppingNSW 1710 Australia Tianjin Key Lab of Intelligent Computing and Novel Software Technology Tianjin University of Technology Tianjin China TianJin Chest Hospital Tianjin University No. 261 Taierzhuang South Road Jinnan District Tianjin China Tianjin Modern Innovative TCM Technology Co. Ltd. Tianjin300392 China The Affiliated Traditional Chinese Medicine Hospital Southwest Medical University Luzhou Sichuan646000 China

At present, anti-cancer drug synergy therapy is one of the most important methods to overcome drug resistance and reduce drug toxicity in cancer treatment. High-throughput screening through deep learning can effectively improve the efficiency of discovering synergistic drugs. Nowadays, most of the existing deep learning algorithms for anti-cancer drug synergy prediction use deep neural networks and can only implicitly perform feature interaction. This study proposes a deep learning algorithm, named MolCross, which combines implicit feature interaction with explicit features to improve the accuracy of prediction of the anti-cancer drug synergy score. MolCross uses a deep autoencoder to extract features from high-dimensional input, uses the drug-specific subnetworks and cross-network to perform implicit feature interaction and explicit feature interaction respectively, and finally uses a synergy prediction network to combine the two feature interaction methods to obtain the final prediction results. We adopted a five-fold cross validation and compared MolCross with other four anti-cancer drug synergy prediction models. The results show that MolCross has better prediction performance than other models. MolCross also has good performance in terms of cross-cell line and cross-tissue type. Existing studies have demonstrated that cancer molecular subtypes have different sensitivities to targeted therapy. In this study, the features of cancer molecular subtype were introduced in the model using an embedding layer in MolCross to explore the effect of cancer molecular subtype on anti-cancer drug synergy. We also found that the cancer molecular subtype is one of the main factors affecting the synergy between drugs. © 2024 Elsevier Ltd

关键词： Forecasting

Active Lighting Recurrence by Parallel Lighting Analogy for Fine-Grained Change Detection

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zhang, Qian Feng, Wei Wan, Liang Tian, Fei-Peng Wang, Xiaowei Tan, Ping School of Computer Science and Technology College of Intelligence and Computing Tianjin University TianJin300350 China State Administration of Cultural Heritage China School of Computer Software College of Intelligence and Computing Tianjin University TianJin300350 China Tianjin Key Lab for Advanced Signal Processing Civil Aviation University of China Dunhuang Research Academy Dunhuang Gansu China School of Computing Science Simon Fraser University Canada

—This paper studies a new problem, namely active lighting recurrence (ALR) that physically relocalizes a light source to reproduce the lighting condition from single reference image for a same scene, which may suffer from fine-grained changes during twice observations. ALR is of great importance for fine-grained visual inspection and change detection, because some phenomena or minute changes can only be clearly observed under particular lighting conditions. Therefore, effective ALR should be able to online navigate a light source toward the target pose, which is challenging due to the complexity and diversity of real-world lighting and imaging processes. To this end, we propose to use the simple parallel lighting as an analogy model and based on Lambertian law to compose an instant navigation ball for this purpose. We theoretically prove the feasibility, i.e., equivalence and convergence, of this ALR approach for realistic near point light source and small near surface light source. Besides, we also theoretically prove the invariance of our ALR approach to the ambiguity of normal and lighting decomposition. The effectiveness and superiority of the proposed approach have been verified by both extensive quantitative experiments and challenging real-world tasks on fine-grained change detection of cultural heritages. We also validate the generality of our approach to non-Lambertian scenes. Copyright © 2020, The Authors. All rights reserved.

关键词： Light sources

DeePMD-kit v3: A Multiple-Backend Framework for Machine Learning Potentials

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zeng, Jinzhe Zhang, Duo Peng, Anyang Zhang, Xiangyu He, Sensen Wang, Yan Liu, Xinzijian Bi, Hangrui Li, Yifan Cai, Chun Zhang, Chengqian Du, Yiming Zhu, Jia-Xin Mo, Pinghui Huang, Zhengtao Zeng, Qiyu Shi, Shaochen Qin, Xuejian Yu, Zhaoxi Luo, Chenxing Ding, Ye Liu, Yun-Pei Shi, Ruosong Wang, Zhenyu Bore, Sigbjørn Løland Chang, Junhan Deng, Zhe Ding, Zhaohan Han, Siyuan Jiang, Wanrun Ke, Guolin Liu, Zhaoqing Lu, Denghui Muraoka, Koki Oliaei, Hananeh Singh, Anurag Kumar Que, Haohui Xu, Weihong Xu, Zhangmancang Zhuang, Yong-Bin Dai, Jiayu Giese, Timothy J. Jia, Weile Xu, Ben York, Darrin M. Zhang, Linfeng Wang, Han School of Artificial Intelligence and Data Science Unversity of Science and Technology of China Hefei China AI for Science Institute Beijing100080 China DP Technology Beijing100080 China Academy for Advanced Interdisciplinary Studies Peking University Beijing100871 China State Key Lab of Processors Institute of Computing Technology Chinese Academy of Sciences Beijing100871 China University of Chinese Academy of Sciences Beijing China Baidu Inc. Beijing China Department of Computer Science University of Toronto TorontoON Canada Department of Chemistry Princeton University PrincetonNJ08540 United States University of Chinese Academy of Sciences Beijing100871 China State Key Laboratory of Physical Chemistry of Solid Surfaces iChEM College of Chemistry and Chemical Engineering Xiamen University Xiamen361005 China College of Integrated Circuits Hunan University Changsha410082 China State Key Laboratory of Advanced Technology for Materials Synthesis and Processing Center for Smart Materials and Device Integration School of Material Science and Engineering Wuhan University of Technology Wuhan430070 China College of Science National University of Defense Technology Changsha410073 China Hunan Key Laboratory of Extreme Matter and Applications National University of Defense Technology Changsha410073 China ByteDance Research Beijing100098 China Ningbo Institute of Materials Technology and Engineering Chinese Academy of Sciences Ningbo315201 China College of Materials Science and Opto-Electronic Technology University of Chinese Academy of Sciences Beijing100049 China Key Laboratory of Theoretical and Computational Photochemistry of Ministry of Education College of Chemistry Beijing Normal University Beijing100875 China Department of Geosciences Princeton University PrincetonNJ08544 United States Department of Applied Physics and Applied Mathematics Columbia University New YorkNY10027 United States IKKEM Fujian Xiamen361005 China Graduate

In recent years, machine learning potentials (MLPs) have become indispensable tools in physics, chemistry, and materials science, driving the development of software packages for molecular dynamics (MD) simulations and related applications. These packages, typically built on specific machine learning frameworks such as TensorFlow, PyTorch, or JAX, face integration challenges when advanced applications demand communication across different frameworks. The previous TensorFlow-based implementation of DeePMD-kit exemplified these limitations. In this work, we introduce DeePMD-kit version 3, a significant update featuring a multi-backend framework that supports TensorFlow, PyTorch, JAX, and PaddlePaddle backends, and demonstrate the versatility of this architecture through the integration of other MLPs packages and of Differentiable Molecular Force Field. This architecture allows seamless backend switching with minimal modifications, enabling users and developers to integrate DeePMD-kit with other packages using different machine learning frameworks. This innovation facilitates the development of more complex and interoperable workflows, paving the way for broader applications of MLPs in scientific research. Copyright © 2025, The Authors. All rights reserved.

关键词： Molecular dynamics

T2-GNN: Graph Neural Networks for Graphs with Incomplete Features and Structure via Teacher-Student Distillation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Huo, Cuiying Jin, Di Li, Yawen He, Dongxiao Yang, Yu-Bin Wu, Lingfei College of Intelligence and Computing Tianjin University Tianjin China State Key Laboratory for Novel Software Technology Nanjing University Nanjing China School of Economics and Management Beijing University of Posts and Telecommunications Beijing China Pinterest New York United States

Graph Neural Networks (GNNs) have been a prevailing technique for tackling various analysis tasks on graph data. A key premise for the remarkable performance of GNNs relies on complete and trustworthy initial graph descriptions (i.e., node features and graph structure), which is often not satisfied since real-world graphs are often incomplete due to various unavoidable factors. In particular, GNNs face greater challenges when both node features and graph structure are incomplete at the same time. The existing methods either focus on feature completion or structure completion. They usually rely on the matching relationship between features and structure, or employ joint learning of node representation and feature (or structure) completion in the hope of achieving mutual benefit. However, recent studies confirm that the mutual interference between features and structure leads to the degradation of GNN performance. When both features and structure are incomplete, the mismatch between features and structure caused by the missing randomness exacerbates the interference between the two, which may trigger incorrect completions that negatively affect node representation. To this end, in this paper we propose a general GNN framework based on teacher-student distillation to improve the performance of GNNs on incomplete graphs, namely T2-GNN. To avoid the interference between features and structure, we separately design feature-level and structure-level teacher models to provide targeted guidance for student model (base GNNs, such as GCN) through distillation. Then we design two personalized methods to obtain well-trained feature and structure teachers. To ensure that the knowledge of the teacher model is comprehensively and effectively distilled to the student model, we further propose a dual distillation mode to enable the student to acquire as much expert knowledge as possible. Extensive experiments on eight benchmark datasets demonstrate the effectiveness and robustness of

关键词： Graph neural networks

Dynamics of Interacting Diseases

学校读者我要写书评

暂无评论

Physical Review X 2014年第4期4卷 041005-041005页

作者： Joaquín Sanz Cheng-Yi Xia Sandro Meloni Yamir Moreno Institute for Biocomputation and Physics of Complex Systems (BIFI) University of Zaragoza Zaragoza 50018 Spain Department of Theoretical Physics University of Zaragoza Zaragoza 50009 Spain Key Laboratory of Computer Vision and System (Ministry of Education) and Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology Tianjin University of Technology Tianjin 300191 People’s Republic of China Complex Networks and Systems Lagrange Lab Institute for Scientific Interchange Turin 10126 Italy

Current modeling of infectious diseases allows for the study of complex and realistic scenarios that go from the population to the individual level of description. However, most epidemic models assume that the spreading process takes place on a single level (be it a single population, a metapopulation system, or a network of contacts). In particular, interdependent contagion phenomena can be addressed only if we go beyond the scheme-one pathogen-one network. In this paper, we propose a framework that allows us to describe the spreading dynamics of two concurrent diseases. Specifically, we characterize analytically the epidemic thresholds of the two diseases for different scenarios and compute the temporal evolution characterizing the unfolding dynamics. Results show that there are regions of the parameter space in which the onset of a disease’s outbreak is conditioned to the prevalence levels of the other disease. Moreover, we show, for the susceptible-infected-susceptible scheme, that under certain circumstances, finite and not vanishing epidemic thresholds are found even at the limit for scale-free networks. For the susceptible-infected-removed scenario, the phenomenology is richer and additional interdependencies show up. We also find that the secondary thresholds for the susceptible-infected-susceptible and susceptible-infected-removed models are different, which results directly from the interaction between both diseases. Our work thus solves an important problem and paves the way toward a more comprehensive description of the dynamics of interacting diseases.

关键词： Complex systems Interdisciplinary physics Statistical physics