[{"data":1,"prerenderedAt":659},["ShallowReactive",2],{"content-query-U9NN2meh61":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":653,"_id":654,"_source":655,"_file":656,"_stem":657,"_extension":658},"/news/zh/2838","zh",false,"","打通科学智能任督二脉 | AI4Science高频模型套件：MindSpore SciAI 0.1发布","2023年9月21日下午，在以“加速行业智能化”为主题的华为全联接大会2023(Huawei Connect2023)昇思MindSpore专题论坛上，昇思MindSpore开源社区正式发布MindSpore AI4Science高频模型套件。","2023-10-20","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2023/10/25/f78fcfee9b134a46940af8418ec30e78.png","news",{"type":14,"children":15,"toc":636},"root",[16,24,29,42,51,56,75,83,88,93,100,105,113,118,125,130,161,166,171,182,191,199,204,214,222,229,239,247,257,267,277,288,298,309,319,330,340,348,353,360,368,373,383,391,396,404,414,419,427,437,442,450,460,468,473,481,486,496,501,509,519,527,537,542,550,555,565,575,583,596,606,611,616,621,626,631],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"打通科学智能任督二脉-ai4science高频模型套件mindspore-sciai-01发布",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":9},{"type":17,"tag":25,"props":30,"children":31},{},[32,34,40],{"type":23,"value":33},"昇思MindSpore长期布局AI4Science（科学智能）领域，不断瞄准行业热点与痛点问题，以其卓越的性能和创新的能力为科学智能领域带来了巨大的助力。MindSpore SciAI 0.1版本内置了60+高频SOTA模型，覆盖物理感知（如PINNs、DeepRitz以及PFNN）和神经算子（如FNO、DeepONet、PDENet）等主流模型，模型覆盖度全球第一；提供了高阶API（一键环境配置、自动模型加载、极简训练微调等），开发者和用户开箱即用。MindSpore SciAI为广大开发者和用户提供了高效、易用的AI4Science科学智能套件。点击",{"type":17,"tag":35,"props":36,"children":37},"strong",{},[38],{"type":23,"value":39},"阅读原文",{"type":23,"value":41},"可前往代码仓下载。",{"type":17,"tag":25,"props":43,"children":44},{},[45],{"type":17,"tag":46,"props":47,"children":50},"img",{"alt":48,"src":49},"image.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025014904.74958760346292149618233285568434:50541024031318:2400:82B314C902AB793FC98D5D6F6E29B774C8B0C57FC48E0D13D392F8C95833B859.png",[],{"type":17,"tag":25,"props":52,"children":53},{},[54],{"type":23,"value":55},"图1 MindSpore SciAI套件架构图",{"type":17,"tag":25,"props":57,"children":58},{},[59],{"type":17,"tag":35,"props":60,"children":61},{},[62],{"type":17,"tag":35,"props":63,"children":64},{},[65,67],{"type":23,"value":66},"1、 ",{"type":17,"tag":35,"props":68,"children":69},{},[70],{"type":17,"tag":35,"props":71,"children":72},{},[73],{"type":23,"value":74},"SOTA模型",{"type":17,"tag":25,"props":76,"children":77},{},[78],{"type":17,"tag":35,"props":79,"children":80},{},[81],{"type":23,"value":82},"1.1、物理感知",{"type":17,"tag":25,"props":84,"children":85},{},[86],{"type":23,"value":87},"物理感知模型是指将物理学中的先验知识（如方程/初边界条件等）融入到神经网络中的一类模型，较为典型的有PINNs（如图2所示）、Deep Ritz以及PFNN等。该类模型的优势：无需生成离散网格；可基于AI框架的自动微分能力进行导数计算，避免数值微分离散误差；天然适用于反问题及数据同化问题；相比数据驱动具有更强的外插能力和更少的样本量。劣势：缺乏网络结构设计指引，奇异性问题学习困难；损失函数包含多项约束，训练难以收敛；物理约束变化时需要重新训练，缺乏泛化性；计算精度和收敛缺乏理论保证。",{"type":17,"tag":25,"props":89,"children":90},{},[91],{"type":23,"value":92},"针对上述问题，学术界和昇思MindSpore提出了一系列改进方案，如自适应激活函数、时间&空间分解、多尺度优化以及自适应加权等。MindSpore SciAI内置了SOTA物理感知模型，应用领域也涵盖了流体、电磁、声、热、固体等众多领域。",{"type":17,"tag":25,"props":94,"children":95},{},[96],{"type":17,"tag":46,"props":97,"children":99},{"alt":48,"src":98},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025014920.49547854883197090216377469758870:50541024031318:2400:DDFBACFB0E70AD3867008323DEEC8FCC281D1EC5FF516E18F28F14BFB652E844.png",[],{"type":17,"tag":25,"props":101,"children":102},{},[103],{"type":23,"value":104},"图2 物理感知模型之PINNs",{"type":17,"tag":25,"props":106,"children":107},{},[108],{"type":17,"tag":35,"props":109,"children":110},{},[111],{"type":23,"value":112},"1.2、神经算子",{"type":17,"tag":25,"props":114,"children":115},{},[116],{"type":23,"value":117},"PINNs等物理感知模型主要求解特定方程，神经算子模型则能够学习无限维函数空间的映射，一次求解整个PDEs族。较为典型的有FNO（如图3所示）、DeepONet以及PDENet等。FNO主要利用傅里叶变换的性质，在傅里叶空间中学习函数之间的映射，然后再将结果转回至物理空间。DeepONet通过“branch net”和“trunk net”两个子网络学习函数之间的映射。神经算子模型在流体、气象等领域有较好的表现，MindSpore SciAI因此也内置了SOTA神经算子模型。",{"type":17,"tag":25,"props":119,"children":120},{},[121],{"type":17,"tag":46,"props":122,"children":124},{"alt":48,"src":123},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025014936.44286956363927692237732124131744:50541024031318:2400:24779D4B6F05A748B8D543916CD2DB9C339CB82E6B7BBA9AA9B77C1A96ADBF02.png",[],{"type":17,"tag":25,"props":126,"children":127},{},[128],{"type":23,"value":129},"图3神经算子模型之FNO",{"type":17,"tag":25,"props":131,"children":132},{},[133],{"type":17,"tag":35,"props":134,"children":135},{},[136],{"type":17,"tag":35,"props":137,"children":138},{},[139,141],{"type":23,"value":140},"2、",{"type":17,"tag":35,"props":142,"children":143},{},[144],{"type":17,"tag":35,"props":145,"children":146},{},[147],{"type":17,"tag":35,"props":148,"children":149},{},[150],{"type":17,"tag":35,"props":151,"children":152},{},[153],{"type":17,"tag":35,"props":154,"children":155},{},[156],{"type":17,"tag":35,"props":157,"children":158},{},[159],{"type":23,"value":160},"总结与展望",{"type":17,"tag":25,"props":162,"children":163},{},[164],{"type":23,"value":165},"昇思MindSpore SciAI 0.1版本正式发布，希望有更多的企业、科研院所携手昇思，共同打造自主创新的AI4Science科学智能套件，加速助力人工智能产业生态繁荣发展。",{"type":17,"tag":25,"props":167,"children":168},{},[169],{"type":23,"value":170},"代码仓地址为：",{"type":17,"tag":25,"props":172,"children":173},{},[174],{"type":17,"tag":175,"props":176,"children":180},"a",{"href":177,"rel":178},"https://gitee.com/mindspore/mindscience/tree/master/SciAI",[179],"nofollow",[181],{"type":23,"value":177},{"type":17,"tag":183,"props":184,"children":186},"h2",{"id":185},"附录",[187],{"type":17,"tag":35,"props":188,"children":189},{},[190],{"type":23,"value":185},{"type":17,"tag":25,"props":192,"children":193},{},[194],{"type":17,"tag":35,"props":195,"children":196},{},[197],{"type":23,"value":198},"附录1、模块及功能",{"type":17,"tag":25,"props":200,"children":201},{},[202],{"type":23,"value":203},"本节主要介绍MindSpore SciAI高阶和基础API。MindSpore SciAI目录结构如下：",{"type":17,"tag":205,"props":206,"children":208},"pre",{"code":207},"├── SciAI\n│   ├── cmake               # 编译脚本\n│   ├── docs                # 文档\n│   │   ├── FAQ        # 常见问题解答\n│   ├── sciai               # MindSpore SciAI主目录\n│   │   ├── architecture   # 神经网络基础模块\n│   │   ├── common          # 通用模块\n│   │   ├── context         # 上下文设置\n│   │   ├── model           # **AI4Science高频模型**\n│   │   ├── operators       # 高阶微分\n│   │   └── utils           # 其他辅助功能\n│   └── tutorial            # 教学模型\n",[209],{"type":17,"tag":210,"props":211,"children":212},"code",{"__ignoreMap":7},[213],{"type":23,"value":207},{"type":17,"tag":25,"props":215,"children":216},{},[217],{"type":17,"tag":35,"props":218,"children":219},{},[220],{"type":23,"value":221},"1.1高阶API",{"type":17,"tag":25,"props":223,"children":224},{},[225],{"type":17,"tag":46,"props":226,"children":228},{"alt":48,"src":227},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025014956.42027150197248344587491510162279:50541024031318:2400:F36EEC2859DAB8ABC89209E79A48E58F2949ABF78854D36AE93BD6CFE496F009.png",[],{"type":17,"tag":230,"props":231,"children":233},"h3",{"id":232},"_12基础api",[234],{"type":17,"tag":35,"props":235,"children":236},{},[237],{"type":23,"value":238},"1.2、基础API",{"type":17,"tag":25,"props":240,"children":241},{},[242],{"type":17,"tag":35,"props":243,"children":244},{},[245],{"type":23,"value":246},"1.2.1、基础API",{"type":17,"tag":25,"props":248,"children":249},{},[250],{"type":17,"tag":35,"props":251,"children":252},{},[253],{"type":17,"tag":46,"props":254,"children":256},{"alt":48,"src":255},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015015.95583183618251957847336859170722:50541024031318:2400:D52848CEE8419D85FD067B7DDFD260A949422D620A1E728A07A0AC18801D07DA.png",[],{"type":17,"tag":258,"props":259,"children":261},"h4",{"id":260},"_122通用模块",[262],{"type":17,"tag":35,"props":263,"children":264},{},[265],{"type":23,"value":266},"1.2.2、通用模块",{"type":17,"tag":25,"props":268,"children":269},{},[270],{"type":17,"tag":35,"props":271,"children":272},{},[273],{"type":17,"tag":46,"props":274,"children":276},{"alt":48,"src":275},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015032.55020701026285310414514088592599:50541024031318:2400:2E3E88741335B9FF8B547ED35DF3B32731B4C0BC197F734EF6C329496ED549AC.png",[],{"type":17,"tag":258,"props":278,"children":280},{"id":279},"_123上下文设置",[281,283],{"type":23,"value":282},"**1.2.3、**",{"type":17,"tag":35,"props":284,"children":285},{},[286],{"type":23,"value":287},"上下文设置",{"type":17,"tag":25,"props":289,"children":290},{},[291],{"type":17,"tag":35,"props":292,"children":293},{},[294],{"type":17,"tag":46,"props":295,"children":297},{"alt":48,"src":296},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015051.58327216481374952868185576970770:50541024031318:2400:F6E15234FF078731B135979C4217DEDAA940A699C498580698CD7B5394F498AC.png",[],{"type":17,"tag":258,"props":299,"children":301},{"id":300},"_124高阶求导函数",[302,304],{"type":23,"value":303},"**1.2.4、**",{"type":17,"tag":35,"props":305,"children":306},{},[307],{"type":23,"value":308},"高阶求导函数",{"type":17,"tag":25,"props":310,"children":311},{},[312],{"type":17,"tag":35,"props":313,"children":314},{},[315],{"type":17,"tag":46,"props":316,"children":318},{"alt":48,"src":317},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015110.38896350990372378761422100637664:50541024031318:2400:BFFD5A0BA9D647638BC3C83FC579F1FABB932128DE916795BB380BDF6E2174BF.png",[],{"type":17,"tag":258,"props":320,"children":322},{"id":321},"_125辅助功能",[323,325],{"type":23,"value":324},"**1.2.5、**",{"type":17,"tag":35,"props":326,"children":327},{},[328],{"type":23,"value":329},"辅助功能",{"type":17,"tag":25,"props":331,"children":332},{},[333],{"type":17,"tag":35,"props":334,"children":335},{},[336],{"type":17,"tag":46,"props":337,"children":339},{"alt":48,"src":338},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015130.01406830879911823433368823789603:50541024031318:2400:2D4DB2D12C6BE8A22960EF847C4968A363EF7CDB93BF125AF28138809FA8F800.png",[],{"type":17,"tag":25,"props":341,"children":342},{},[343],{"type":17,"tag":35,"props":344,"children":345},{},[346],{"type":23,"value":347},"附录2、网络模型库",{"type":17,"tag":25,"props":349,"children":350},{},[351],{"type":23,"value":352},"MindSpore SciAI提供了丰富的AI4Science高频模型，下图汇总了当前已实现的网络模型及其对应领域。",{"type":17,"tag":25,"props":354,"children":355},{},[356],{"type":17,"tag":46,"props":357,"children":359},{"alt":48,"src":358},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015148.02536702764311783433430866251734:50541024031318:2400:A80094C24A13890FA5B9B0789C4BE2C4FC637513D4C788345071A207D88293DE.png",[],{"type":17,"tag":25,"props":361,"children":362},{},[363],{"type":17,"tag":35,"props":364,"children":365},{},[366],{"type":23,"value":367},"附录3、使用方式",{"type":17,"tag":25,"props":369,"children":370},{},[371],{"type":23,"value":372},"MindSpore SciAI提供两种模型训练和评估方式，为开发者提供灵活且简洁的使用方式和开发接口。",{"type":17,"tag":25,"props":374,"children":375},{},[376,378],{"type":23,"value":377},"**3.1、**",{"type":17,"tag":35,"props":379,"children":380},{},[381],{"type":23,"value":382},"高阶API启动（推荐）",{"type":17,"tag":25,"props":384,"children":385},{},[386],{"type":17,"tag":35,"props":387,"children":388},{},[389],{"type":23,"value":390},"3.1.1、使用AutoModel训练、微调模型",{"type":17,"tag":25,"props":392,"children":393},{},[394],{"type":23,"value":395},"根据AutoModel.from_pretrained接口获取已支持的网络模型。使用AutoModel.train实现模型的训练，并且在训练前，可使用AutoModel.update_config调整训练参数或加载.ckpt文件实现模型微调。AutoModel.update_config可选参数依赖于模型类型，具体可参考网络模型库中“MindSpore实现与网络参数链接”。",{"type":17,"tag":205,"props":397,"children":399},{"code":398},"from sciai.model import AutoModel\n\n# 获取`cpinns`网络模型\nmodel = AutoModel.from_pretrained(\"cpinns\")\n# 使用默认参数训练网络，生成的图片、数据与日志将保存至用户的执行目录中\nmodel.train()\n# 或者加载`.ckpt`文件\nmodel.update_config(load_ckpt=True, load_ckpt_path=\"./checkpoints/your_file.ckpt\", epochs=500)\n# 基于新加载的参数，继续训练模型，实现微调\nmodel.train()\n",[400],{"type":17,"tag":210,"props":401,"children":402},{"__ignoreMap":7},[403],{"type":23,"value":398},{"type":17,"tag":25,"props":405,"children":406},{},[407,409],{"type":23,"value":408},"**3.1.2、**",{"type":17,"tag":35,"props":410,"children":411},{},[412],{"type":23,"value":413},"使用AutoModel评估模型",{"type":17,"tag":25,"props":415,"children":416},{},[417],{"type":23,"value":418},"用户可以使用AutoModel.evaluate评估训练结果。该接口将默认加载SciAI模型库中提供的.ckpt文件用于评估，用户也可以调用model.update_config接口自定义加载的文件。",{"type":17,"tag":205,"props":420,"children":422},{"code":421},"from sciai.model import AutoModel\n\n# 获取`cpinns`网络模型\nmodel = AutoModel.from_pretrained(\"cpinns\")\n# 加载网络默认的ckpt文件，评估网络模型\nmodel.evaluate()\n# 自定义加载ckpt文件\nmodel.update_config(load_ckpt=True, load_ckpt_path=\"./checkpoints/your_file.ckpt\")\n# 评估网络模型\nmodel.evaluate()\n",[423],{"type":17,"tag":210,"props":424,"children":425},{"__ignoreMap":7},[426],{"type":23,"value":421},{"type":17,"tag":25,"props":428,"children":429},{},[430,432],{"type":23,"value":431},"**3.1.3、**",{"type":17,"tag":35,"props":433,"children":434},{},[435],{"type":23,"value":436},"恢复参数文件与数据集",{"type":17,"tag":25,"props":438,"children":439},{},[440],{"type":23,"value":441},"用户可能会遇到自动下载的模型参数.ckpt文件或数据集被误删的情况，可以通过如下方式更新参数，强制重新下载。",{"type":17,"tag":205,"props":443,"children":445},{"code":444},"model.update_config(force_download=True)\n",[446],{"type":17,"tag":210,"props":447,"children":448},{"__ignoreMap":7},[449],{"type":23,"value":444},{"type":17,"tag":25,"props":451,"children":452},{},[453,455],{"type":23,"value":454},"**3.2、**",{"type":17,"tag":35,"props":456,"children":457},{},[458],{"type":23,"value":459},"脚本接口",{"type":17,"tag":25,"props":461,"children":462},{},[463],{"type":17,"tag":35,"props":464,"children":465},{},[466],{"type":23,"value":467},"3.2.1、环境准备",{"type":17,"tag":25,"props":469,"children":470},{},[471],{"type":23,"value":472},"首先使用如下命令直接克隆整个仓库并准备环境变量，运行文件.env会将SciAI项目目录添加至环境变量PYTHONPATH中。",{"type":17,"tag":205,"props":474,"children":476},{"code":475},"git clone https://gitee.com/mindspore/mindscience\nsource ./mindscience/SciAI/.env\n",[477],{"type":17,"tag":210,"props":478,"children":479},{"__ignoreMap":7},[480],{"type":23,"value":475},{"type":17,"tag":25,"props":482,"children":483},{},[484],{"type":23,"value":485},"完成后，可以参考网络模型库中各个模型README_CN.md 的快速开始章节，使用源码进行训练与推理。",{"type":17,"tag":25,"props":487,"children":488},{},[489,491],{"type":23,"value":490},"**3.2.2、**",{"type":17,"tag":35,"props":492,"children":493},{},[494],{"type":23,"value":495},"网络模型训练",{"type":17,"tag":25,"props":497,"children":498},{},[499],{"type":23,"value":500},"使用训练脚本train.py进行网络模型训练。",{"type":17,"tag":205,"props":502,"children":504},{"code":503},"cd ./mindscience/SciAI/sciai/model/cpinns/\npython ./train.py [--parameters]\n",[505],{"type":17,"tag":210,"props":506,"children":507},{"__ignoreMap":7},[508],{"type":23,"value":503},{"type":17,"tag":25,"props":510,"children":511},{},[512,514],{"type":23,"value":513},"**3.2.3、**",{"type":17,"tag":35,"props":515,"children":516},{},[517],{"type":23,"value":518},"网络模型加载.ckpt文件微调",{"type":17,"tag":205,"props":520,"children":522},{"code":521},"python ./train.py --load_ckpt true --load_ckpt_path {your_ckpt_file.ckpt} \n[--parameters]\n",[523],{"type":17,"tag":210,"props":524,"children":525},{"__ignoreMap":7},[526],{"type":23,"value":521},{"type":17,"tag":25,"props":528,"children":529},{},[530,532],{"type":23,"value":531},"**3.2.4、**",{"type":17,"tag":35,"props":533,"children":534},{},[535],{"type":23,"value":536},"网络模型评估",{"type":17,"tag":25,"props":538,"children":539},{},[540],{"type":23,"value":541},"使用脚本eval.py对已完成训练的网络模型进行评估。",{"type":17,"tag":205,"props":543,"children":545},{"code":544},"python ./eval.py [--parameters]\n",[546],{"type":17,"tag":210,"props":547,"children":548},{"__ignoreMap":7},[549],{"type":23,"value":544},{"type":17,"tag":25,"props":551,"children":552},{},[553],{"type":23,"value":554},"可配置参数[--parameters]包括学习率、训练周期、数据读取保存路径、checkpoints文件加载路径等，具体请参考网络模型库中“昇思MindSpore实现与网络参数链接”。",{"type":17,"tag":25,"props":556,"children":557},{},[558,560],{"type":23,"value":559},"**3.2.5、**",{"type":17,"tag":35,"props":561,"children":562},{},[563],{"type":23,"value":564},"cpinns计算结果如下",{"type":17,"tag":25,"props":566,"children":567},{},[568],{"type":17,"tag":35,"props":569,"children":570},{},[571],{"type":17,"tag":46,"props":572,"children":574},{"alt":48,"src":573},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231025015213.01210655122913931253433687594171:50541024031318:2400:1287EB4497B8BECD70317DB96D86CEC583AF06A0601C33671DABD8461546B6BB.png",[],{"type":17,"tag":183,"props":576,"children":578},{"id":577},"参考文献",[579],{"type":17,"tag":35,"props":580,"children":581},{},[582],{"type":23,"value":577},{"type":17,"tag":25,"props":584,"children":585},{},[586,588,594],{"type":23,"value":587},"[1] ",{"type":17,"tag":175,"props":589,"children":592},{"href":590,"rel":591},"https://gitee.com/mindspore/mindscience/tree/master",[179],[593],{"type":23,"value":590},{"type":23,"value":595},".",{"type":17,"tag":25,"props":597,"children":598},{},[599,601],{"type":23,"value":600},"[2] ",{"type":17,"tag":175,"props":602,"children":604},{"href":177,"rel":603},[179],[605],{"type":23,"value":177},{"type":17,"tag":25,"props":607,"children":608},{},[609],{"type":23,"value":610},"[3] Raissi M, Perdikaris P, Karniadakis G E. Physics informed deep learning (part i): Data-driven solutions of nonlinear partial differential equations[J]. arXiv preprint arXiv:1711.10561, 2017.",{"type":17,"tag":25,"props":612,"children":613},{},[614],{"type":23,"value":615},"[4] Yu B. The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems[J]. Communications in Mathematics and Statistics, 2018, 6(1): 1-12.",{"type":17,"tag":25,"props":617,"children":618},{},[619],{"type":23,"value":620},"[5] Sheng H, Yang C. PFNN: A penalty-free neural network method for solving a class of second-order boundary-value problems on complex geometries[J]. Journal of Computational Physics, 2021, 428: 110085.",{"type":17,"tag":25,"props":622,"children":623},{},[624],{"type":23,"value":625},"[6] Li Z, Kovachki N, Azizzadenesheli K, et al. Fourier neural operator for parametric partial differential equations[J]. arXiv preprint arXiv:2010.08895, 2020.",{"type":17,"tag":25,"props":627,"children":628},{},[629],{"type":23,"value":630},"[7] Lu L, Jin P, Karniadakis G E. Deeponet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators[J]. arXiv preprint arXiv:1910.03193, 2019.",{"type":17,"tag":25,"props":632,"children":633},{},[634],{"type":23,"value":635},"[8] Long Z, Lu Y, Ma X, et al. Pde-net: Learning pdes from data[C]//International conference on machine learning. PMLR, 2018: 3208-3216.",{"title":7,"searchDepth":637,"depth":637,"links":638},4,[639,652],{"id":185,"depth":640,"text":185,"children":641},2,[642],{"id":232,"depth":643,"text":238,"children":644},3,[645,646,648,650],{"id":260,"depth":637,"text":266},{"id":279,"depth":637,"text":647},"**1.2.3、**上下文设置",{"id":300,"depth":637,"text":649},"**1.2.4、**高阶求导函数",{"id":321,"depth":637,"text":651},"**1.2.5、**辅助功能",{"id":577,"depth":640,"text":577},"markdown","content:news:zh:2838.md","content","news/zh/2838.md","news/zh/2838","md",1776506073923]