captaintsao / mydl Goto Github PK
View Code? Open in Web Editor NEWThis project forked from myazi/mydl
This project forked from myazi/mydl
1.DNN 可设置网络结构参数,层数,神经元个数,激活函数 -------------------------------------------------------- 2.初始化: Random,"he","arXiv" 3.方差/偏差: Regularization, Dropout 4 随机采样: Stochastic gradient descent,mini-batch 5 优化方法: Gradient Descent,Momentum Gradient Descent,Adam 6 矩阵库更新: 主要考虑内存分配问题,但存在申请内存失败问题 -------------------------------------------------------------------- 1 深度神经网络: C++/python/TF 实现 2 卷积神经网络: python/TF 实现 3 序列模型: python/TF 实现 (RNN GRU LSTM) 4 DeepLearing学习笔记: 提纲 1神经网络构建 1.1神经网络基础 逻辑回归模型,Softmax,矩阵形式表示 1.2 神经网络的前向传递和反向传播的数学推导 线性,激活函数,损失函数,需要注意的是数据表示,维度问题 2 改善神经网络 2.1 防止过拟合 提前结束, 增加训练样本, 正则项, Dropout 2.2 梯度消失问题 初始化,梯度剪辑,激活函数, Batch-Norm,Res-net,LSTM 2.3 优化方法(加速 学习率) Mini-batch,动量的梯度下降,RMS-prop,Adam 2.4 梯度检验, 调参(网络结构:层,神经元个数;学习率: ;优化方法: ;正则化项: ;Dropout;Mini-batch size;迭代次数;初始化方式) 3 结构化机器学习 3.1 快速搭建好第一个系统:框架使用,开源使用, 迁移学习,数据预处理 3.2 误差分析:偏差和方差分析,正交化调试,控制变量法调试 1 偏差问题,调整网络架构,学习率,优化方法,增加数据集 2 方差问题(过拟合),正则化,Dropout,增加训练集 3 真实数据场景差 真实数据与训练测试数据分布不一致 3.3 单一评价标准,如F1值结合了精度和召回率 4 卷积神经网络 4.1 卷积,池化,填充,全连接概念以及超参数和参数(参数共享,特征提取**) 4.2 CNN构建的流程以及数学表示形式,需要注意的是,数据表示,维度问题 1 卷积,Relu,池化,向量化,全连接,Softmax层,前向传播 2 卷积,池化,全连接,Softmax层,链式求导 4.3 经典的CNN架构 LeNet-5,AlexNet,VGG(卷积核大小统一化),Resnet(解决梯度消失问题),Inception(可学习卷积核大小),迁移学习构建初始网络架构 5 序列模型 5.1 序列模型架构:many-to-many(实体识别,机器翻译), many-to-one(情感分类),one-to-many(语音生成),one-to-one 5.2 网络架构:RNN,GRU,LSTM,B-RNN,D-RNN 前向反向数学推导,需要注意的是数据表示,维度问题 5.3 Embeding:上下文关系学习词向量(*),skipgram,负采用,Glove词向量,去偏见,注意力机制(上下文加权)
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.