https://gitee.com/zzhzwh/ZJU_Machine_Learning/tree/main

多层神经网络

image-20211125171453008

image-20211125171645012

image-20211125171900902

image-20211125171933566

image-20211125172020398

image-20211125171954278

image-20211125172112926

image-20211125172127396

image-20211125172256911

image-20211125172309111

image-20211125172459031

image-20211125172613589

image-20211125172653488image-20211125172659102image-20211125172710975

image-20211125172718438

image-20211125172857109

image-20211125172907078

image-20211125173158239

image-20211125173220982

image-20211125173313831

image-20211125173349192

梯度下降

image-20211125173422174

image-20211125173526846

image-20211125173637807

image-20211125173653264

image-20211125173726493

image-20211125173745415

image-20211125173753583

后向传播

image-20211125173955935

image-20211125174011863

image-20211125174028070

image-20211125174109422

image-20211125174125861

image-20211125174135414

image-20211125174215550

image-20211125174249399

image-20211125174344209

image-20211125174438182

image-20211125174527256

image-20211125174542421

image-20211125174548318

一般的后向

image-20211125174750367

image-20211125174804343

image-20211125174833185

image-20211125174911357

image-20211125174927368

image-20211125174957271

image-20211125175015837

image-20211125175053479

image-20211125175058853

image-20211125175106936

image-20211125175202895

后向传播算法的应用和改进

非线性改进

image-20211125175352750

image-20211125175358551

阶跃函数在0处没有导数、

image-20211125175423438

image-20211125175452360

image-20211125175531206

image-20211125175546745

目标函数第二个改进----输出不止一个

image-20211125175701230

原来的

image-20211125175734713

改进的

image-20211125175756113

image-20211125175808349

image-20211125175815431

image-20211125175854775

image-20211125175917160

image-20211125175928951

image-20211125175934283

image-20211125175946886

原来的梯度下降缓慢的运行，而采用随机梯度

image-20211125180021744

image-20211125180100584

image-20211125180113024

image-20211125180138237

人工神经网络的建议

image-20211125181441391

image-20211125181447823

image-20211125181516414

image-20211125181532702

image-20211125181609016

image-20211125181616271

image-20211125181731272

image-20211125181741310

image-20211125181850786

image-20211125181918825

image-20211125181941536

image-20211125182011679

image-20211125182055407

image-20211125182104840

image-20211125182132768

image-20211125182139322

多层神经网络的劣势—进入低潮

image-20211125182314352

image-20211125182340494

image-20211125182354001

image-20211125182457417

深度学习

image-20211125182623043

自编码器

image-20211125202243710

image-20211125202405376

image-20211125202433642

image-20211125202524943

image-20211125202531396

image-20211125202559540

image-20211125202708423

CNN卷积神经网络

image-20211125204752282

image-20211125204800661

求卷积核

image-20211125204908340

image-20211125205013349

image-20211125205023651

image-20211125205029388

image-20211125205141732

image-20211125205213244

image-20211125205219645

image-20211125205355751

image-20211125205403807

image-20211125205443244

image-20211125205449245

步长2 2

image-20211125205502581

image-20211125205521301

降采样

image-20211125205554244

image-20211125205731679

image-20211125205742933

LENET的参数个数

image-20211125205903655

image-20211125210037316

image-20211125210251948

image-20211125210045515

AlexNet

image-20211125210207751

image-20211125210227032

image-20211125210312460

image-20211125210348653

使用RElu函数：促进深度神经网络的收敛
最大池化

image-20211125210503028

image-20211125210520684

随即丢弃

image-20211125210546637

image-20211125210606182

数据扩增

image-20211125210640483

GPU训练

image-20211125210712941

编程工具

image-20211125211538525

image-20211125211613996

pytorch

image-20211125211107264

image-20211125211144453

image-20211125211211996

image-20211125211227452

image-20211125211304247

image-20211125211320567

image-20211125211422475

TensorFlow

caffe

image-20211125213034117

image-20211125213050197

image-20211125213059870

image-20211125212915336

image-20211125213122401

人脸识别

image-20211125213201910

image-20211125213226620

image-20211125213237685

损失函数

image-20211125213307173

image-20211125213325544

image-20211125213435926

image-20211125213448158

image-20211125213516376

image-20211125213540807

人脸检测

image-20211125213737109

图像识别

image-20211125213941048

目标定位

image-20211125214053610

目标检测的3种

image-20211125214116626

image-20211125214131574

image-20211125214143409

image-20211125214200302

image-20211125214214389

image-20211125214241327

image-20211125214325951

image-20211125214633374

image-20211125214643774

YOLo网络 you only look once

image-20211125214816043

image-20211125214824550

image-20211125214842950

image-20211125214848797

image-20211125214903374

image-20211125214909023

MTCNN

image-20211125214942293

image-20211125214954283

image-20211125215006383

image-20211125215022349

语义分割

image-20211125215138007

image-20211125215144513

image-20211125215421243

image-20211125215443894

image-20211125215452824

时间序列的深度学习模型

image-20211125215557102

image-20211125215634630

image-20211125215648732

RNN

image-20211125215718397

image-20211125215730334

image-20211125215902775

image-20211125215924504

image-20211125220003813

image-20211125220011561

image-20211125220030579

image-20211125220045111

image-20211125220057950

image-20211125220112705

image-20211125220129032

image-20211125220154118

image-20211125220227025

image-20211125220236360

image-20211125220244319

image-20211125220257013

image-20211125220310424

LSTN

image-20211125220333775

image-20211125215658308

GAN

image-20211125220513512

博弈论

image-20211125220526910

image-20220301163935765

image-20211125220632051

image-20211125220638857

人脸生成

image-20211125220720377

侧脸转正

image-20211125220736961

图像翻译

image-20211125220747415

image-20211125220757616

image-20211125220830781

强化学习

image-20211125221003318

image-20211125221020085

image-20211125221030421

image-20211125221109303

image-20211125221229511

# 机器学习 # 深度学习 # 人工智能
Logo
魔乐社区

魔乐社区（Modelers.cn) 是一个中立、公益的人工智能社区，提供人工智能工具、模型、数据的托管、展示与应用协同服务，为人工智能开发及爱好者搭建开放的学习交流平台。社区通过理事会方式运作，由全产业链共同建设、共同运营、共同享有，推动国产AI生态繁荣发展。

更多推荐

cover

替你试过了，消费级显卡可以跑的开源文生图SOTA模型，顶级渲染、高密度文本绘图

avatar 魔乐社区
cover

量化挑战赛冠军专访：4小时啃下W4A8量化，我靠的是这些经验

avatar 魔乐社区

小参数・大码力・易部署 | Qwen3.6-27B上线魔乐社区，基于昇腾的部署教程来了

继一周前模型开源发布后，千问再度开源Qwen3.6-27B —— 一个拥有270亿参数的稠密多模态模型，也是社区呼声最高的模型规格。Qwen3.6-27B 依然支持多模态思考与非思考模式，在智能体编程方面达到了旗舰级表现，全面超越前代开源旗舰 Qwen3.5-397B-A17B（总参数397B / 激活参数17B的MoE模型）。作为稠密架构，它无需MoE路由即可部署，是开发者在实用、可广泛部署规模

avatar 魔乐社区