V
主页
京东 11.11 红包
【学术教程】张铎:Transferable, Pretrainable and High-capacity Interatomic Deep...
发布人
【学术教程】张铎:Transferable, Pretrainable and High-capacity Interatomic Deep Potential over Periodic Table of Elements Machine learning assisted modeling of the inter-atomic potential energy surface (PES) is revolutionizing the field of molecular simulation. With the accumulation of high-quality electronic structure data, a model that can be pretrained on all available data and finetuned on downstream tasks with a small additional effort would bring the field to a new stage. Here we propose DPA-1, a Deep Potential model with a novel attention mechanism, which is highly effective for representing the conformation and chemical spaces of atomic systems and learning the PES. We tested DPA-1 on a number of systems and observed superior performance compared with existing benchmarks. When pretrained on large-scale datasets containing 56 elements, DPA-1 can be successfully applied to various downstream tasks with a great improvement of sample efficiency. Surprisingly, for different elements, the learned type embedding parameters form a spiral in the latent space and have a natural correspondence with their positions on the periodic table, showing interesting interpretability of the pretrained DPA-1 model.
打开封面
下载高清视频
观看高清视频
视频下载器
【学术教程】郭资政:High Performance Computing in Chip Design
【学术教程】施朱鸣:An overview of auction problems
【学术教程】吉如一:"AC" Automaton -- The Road to Automatic Algorithm Design
【学术教程】邱天异:Value Alignment: History, Frontiers, and Open Problems
[AAAI 2023] Tracking and Reconstructing Hand Object Interactions from...
对话John Hopcroft教授:新国际环境下的高等教育与学术人生
【学术教程】陈旭雯:Methods for thin-shell soft body simulation
[ICRA 2023] RLAfford: End-to-End Affordance Learning for Robotic Manipulation
【学术教程】林汇平:Quantum Error Correction Codes: a Classical View
北京大学李彤阳-《量子计算》第四节:通用量子门与Deutsch-Jozsa算法
【前沿研究课程】第二讲:Bayesian and Contextual Bandit
【亮点报告】耿浩然:PartManip: Learning Cross-Category Generalizable Part Manipulation...
【亮点报告】罗润冬:Rethinking the Effect of Data Augmentation in Adversarial...
【前沿研究课程】第一讲:Introduction to Multi-Armed Bandits
【亮点报告】张家梁:DexGraspNet: A Large-Scale Robotic Dexterous Grasp Dataset for...
【亮点报告】宁川若:Learning Foresightful Dense Visual Affordance for Deformable...
【学术教程】段志健:Deep Learning-Aided Auction Design
程宽博士:RFF方法在核方法中的上下界分析
北京大学李彤阳-《算法分析和复杂性理论》第四节:Greedy algorithms: Interval scheduling, Huffman codes...
【IJTCS 2020】李建 On Generalization and Implicit Bias of Gradient Methods in Deep L
【亮点报告】王泽州:Electrode: Accelerating Distributed Protocols with eBPF
【IJTCS 2020】蔡天乐 Towards Understanding Optimization of Deep Learning
【前沿研究课程】第三讲:SoK: Consensus in the Age of Blockchains
【亮点报告】耿逸然:RLAfford: End-to-End Affordance Learning for Robotic Manipulation
北京大学李彤阳-《量子计算》第十五节:哈密顿量模拟
[SIGGRAPH 2020] Skeleton-Aware Networks for Deep Motion Retargeting
北京大学李彤阳-《算法分析和复杂性理论》第四节:Greedy algorithms: Interval scheduling, Huffman codes...
【亮点报告】施朱鸣:Dynamic Budget Throttling in Repeated Second-Price Auctions
【亮点报告】黄蔚尧:Omni6DPose: A Benchmark and Model for Universal 6D Object Pose Estimat
[NeurIPS 2023] Where2Explore:让机器人轻松应对未见过的物体挑战
【亮点报告】严汨:MaskClustering for Open-Vocabulary 3D Instance Segmentation
【亮点报告】王颖:Sybil-Proof Diffusion Auction in Social Networks
【IJTCS 2020】Nick Gravin Envy-freeness up to Any Item with High Nash Welfare
北京大学李彤阳-《算法分析和复杂性理论》第十三节:Approximation algorithms(上)
【讲座回放】Dr. Argyrios Deligkas: A Polynomial-Time Algorithm for 1/3-Approximate...
北京大学李彤阳-《算法分析和复杂性理论》第十二节:NP-Completeness(中)
[NeurIPS 2023] 认知层级下的群体动作预测
【讲座回放】黄浩强:Fréchet Distance 的次平方时间算法
北京大学李彤阳-《算法分析和复杂性理论》第七节(下)
北京大学李彤阳-《算法分析和复杂性理论》第十节(下)