From f8a43f1d6eca4b1840160316cce902a47c1f0dff Mon Sep 17 00:00:00 2001 From: Aston Zhang Date: Mon, 2 Jul 2018 19:12:58 +0000 Subject: [PATCH] reorg dl basics --- chapter_deep-learning-basics/dropout-gluon.md | 2 +- .../{dropout-scratch.md => dropout.md} | 2 +- chapter_deep-learning-basics/index.md | 12 ++++++------ .../{kaggle-gluon-kfold.md => kaggle-house-price.md} | 0 ...hallow-model.md => linear-regression-concepts.md} | 2 +- .../linear-regression-gluon.md | 2 +- .../linear-regression-scratch.md | 2 +- .../{multi-layer.md => mlp-concepts.md} | 2 +- chapter_deep-learning-basics/mlp-gluon.md | 2 +- chapter_deep-learning-basics/mlp-scratch.md | 2 +- chapter_deep-learning-basics/reg-gluon.md | 2 +- .../{reg-scratch.md => reg.md} | 2 +- ...ssification.md => softmax-regression-concepts.md} | 2 +- .../softmax-regression-gluon.md | 2 +- .../softmax-regression-scratch.md | 2 +- 15 files changed, 19 insertions(+), 19 deletions(-) rename chapter_deep-learning-basics/{dropout-scratch.md => dropout.md} (99%) rename chapter_deep-learning-basics/{kaggle-gluon-kfold.md => kaggle-house-price.md} (100%) rename chapter_deep-learning-basics/{shallow-model.md => linear-regression-concepts.md} (99%) rename chapter_deep-learning-basics/{multi-layer.md => mlp-concepts.md} (99%) rename chapter_deep-learning-basics/{reg-scratch.md => reg.md} (99%) rename chapter_deep-learning-basics/{classification.md => softmax-regression-concepts.md} (99%) diff --git a/chapter_deep-learning-basics/dropout-gluon.md b/chapter_deep-learning-basics/dropout-gluon.md index 9d63c9c..ebc571a 100644 --- a/chapter_deep-learning-basics/dropout-gluon.md +++ b/chapter_deep-learning-basics/dropout-gluon.md @@ -1,4 +1,4 @@ -# 丢弃法——使用Gluon +# 丢弃法的Gluon实现 本节中，我们将上一节的实验代码用Gluon实现一遍。你会发现代码将精简很多。 diff --git a/chapter_deep-learning-basics/dropout-scratch.md b/chapter_deep-learning-basics/dropout.md similarity index 99% rename from chapter_deep-learning-basics/dropout-scratch.md rename to chapter_deep-learning-basics/dropout.md index dd2c455..c478b16 100644 --- a/chapter_deep-learning-basics/dropout-scratch.md +++ b/chapter_deep-learning-basics/dropout.md @@ -1,4 +1,4 @@ -# 丢弃法——从零开始 +# 丢弃法 除了前两节介绍的权重衰减以外，深度学习模型常常使用丢弃法（dropout）来应对过拟合问题。丢弃法有一些不同的变体。本节中提到的丢弃法特指倒置丢弃法（inverted dropout）。它被广泛使用于深度学习。 diff --git a/chapter_deep-learning-basics/index.md b/chapter_deep-learning-basics/index.md index 23d47ad..d1fb26c 100644 --- a/chapter_deep-learning-basics/index.md +++ b/chapter_deep-learning-basics/index.md @@ -7,21 +7,21 @@ .. toctree:: :maxdepth: 2 - shallow-model + linear-regression-concepts linear-regression-scratch linear-regression-gluon - classification + softmax-regression-concepts softmax-regression-scratch softmax-regression-gluon - multi-layer + mlp-concepts mlp-scratch mlp-gluon underfit-overfit - reg-scratch + reg reg-gluon - dropout-scratch + dropout dropout-gluon backprop - kaggle-gluon-kfold + kaggle-house-price ``` diff --git a/chapter_deep-learning-basics/kaggle-gluon-kfold.md b/chapter_deep-learning-basics/kaggle-house-price.md similarity index 100% rename from chapter_deep-learning-basics/kaggle-gluon-kfold.md rename to chapter_deep-learning-basics/kaggle-house-price.md diff --git a/chapter_deep-learning-basics/shallow-model.md b/chapter_deep-learning-basics/linear-regression-concepts.md similarity index 99% rename from chapter_deep-learning-basics/shallow-model.md rename to chapter_deep-learning-basics/linear-regression-concepts.md index 85aef45..9a41b92 100644 --- a/chapter_deep-learning-basics/shallow-model.md +++ b/chapter_deep-learning-basics/linear-regression-concepts.md @@ -1,4 +1,4 @@ -# 单层神经网络 +# 线性回归的概念 在本章的前几节，让我们重温一些经典的浅层模型，例如线性回归和Softmax回归。前者适用于回归问题：模型最终输出是一个连续值，例如房价；后者适用于分类问题：模型最终输出是一个离散值，例如图片的类别。这两种浅层模型本质上都是单层神经网络。它们涉及到的概念和技术对大多数深度学习模型来说同样适用。 diff --git a/chapter_deep-learning-basics/linear-regression-gluon.md b/chapter_deep-learning-basics/linear-regression-gluon.md index 3d06c62..9198ef8 100644 --- a/chapter_deep-learning-basics/linear-regression-gluon.md +++ b/chapter_deep-learning-basics/linear-regression-gluon.md @@ -1,4 +1,4 @@ -# 线性回归——使用Gluon +# 线性回归的Gluon实现 随着深度学习框架的发展，开发深度学习应用变得越来越便利。实践中，我们通常可以用比上一节中更简洁的代码来实现相同模型。本节中，我们将介绍如何使用MXNet提供的Gluon接口更方便地实现线性回归的训练。 diff --git a/chapter_deep-learning-basics/linear-regression-scratch.md b/chapter_deep-learning-basics/linear-regression-scratch.md index 54bb984..072ac87 100644 --- a/chapter_deep-learning-basics/linear-regression-scratch.md +++ b/chapter_deep-learning-basics/linear-regression-scratch.md @@ -1,4 +1,4 @@ -# 线性回归——从零开始 +# 线性回归的从零开始实现 在了解了线性回归的背景知识之后，现在我们可以动手实现它了。 尽管强大的深度学习框架可以减少大量重复性工作，但若过于依赖它提供的便利，我们就会很难深入理解深度学习是如何工作的。因此，本节将介绍如何只利用NDArray和`autograd`来实现一个线性回归的训练。 diff --git a/chapter_deep-learning-basics/multi-layer.md b/chapter_deep-learning-basics/mlp-concepts.md similarity index 99% rename from chapter_deep-learning-basics/multi-layer.md rename to chapter_deep-learning-basics/mlp-concepts.md index 25900dd..a5fc513 100644 --- a/chapter_deep-learning-basics/multi-layer.md +++ b/chapter_deep-learning-basics/mlp-concepts.md @@ -1,4 +1,4 @@ -# 多层神经网络 +# 多层感知机的概念 我们已经介绍了包括线性回归和Softmax回归在内的单层神经网络。本节中，我们将以多层感知机（multilayer perceptron，简称MLP）为例，介绍多层神经网络的概念。 diff --git a/chapter_deep-learning-basics/mlp-gluon.md b/chapter_deep-learning-basics/mlp-gluon.md index ed5078c..0659f4a 100644 --- a/chapter_deep-learning-basics/mlp-gluon.md +++ b/chapter_deep-learning-basics/mlp-gluon.md @@ -1,4 +1,4 @@ -# 多层感知机——使用Gluon +# 多层感知机的Gluon实现 下面我们使用Gluon来实现上一节中的多层感知机。首先我们导入所需的包或模块。 diff --git a/chapter_deep-learning-basics/mlp-scratch.md b/chapter_deep-learning-basics/mlp-scratch.md index 7d80ab4..87c46cd 100644 --- a/chapter_deep-learning-basics/mlp-scratch.md +++ b/chapter_deep-learning-basics/mlp-scratch.md @@ -1,4 +1,4 @@ -# 多层感知机——从零开始 +# 多层感知机的从零开始实现 我们已经从上一章里了解了多层感知机的原理。下面，我们一起来动手实现一个多层感知机。首先导入实现所需的包或模块。 diff --git a/chapter_deep-learning-basics/reg-gluon.md b/chapter_deep-learning-basics/reg-gluon.md index 269e3fb..40846e2 100644 --- a/chapter_deep-learning-basics/reg-gluon.md +++ b/chapter_deep-learning-basics/reg-gluon.md @@ -1,4 +1,4 @@ -# 权重衰减——使用Gluon +# 权重衰减的Gluon实现 本节将介绍如何使用Gluon实现上一节介绍的权重衰减。首先导入实验所需的包或模块。 diff --git a/chapter_deep-learning-basics/reg-scratch.md b/chapter_deep-learning-basics/reg.md similarity index 99% rename from chapter_deep-learning-basics/reg-scratch.md rename to chapter_deep-learning-basics/reg.md index 06fe668..b906594 100644 --- a/chapter_deep-learning-basics/reg-scratch.md +++ b/chapter_deep-learning-basics/reg.md @@ -1,4 +1,4 @@ -# 权重衰减——从零开始 +# 权重衰减 上一节中我们观察了过拟合现象，即模型的训练误差远小于它在测试数据集上的误差。本节将介绍应对过拟合问题的常用方法：权重衰减。 diff --git a/chapter_deep-learning-basics/classification.md b/chapter_deep-learning-basics/softmax-regression-concepts.md similarity index 99% rename from chapter_deep-learning-basics/classification.md rename to chapter_deep-learning-basics/softmax-regression-concepts.md index ab65fe3..f633c66 100644 --- a/chapter_deep-learning-basics/classification.md +++ b/chapter_deep-learning-basics/softmax-regression-concepts.md @@ -1,4 +1,4 @@ -# 分类模型 +# Softmax回归的概念 前几节介绍的线性回归模型适用于输出为连续值的情景，例如输出为房价。在其他情景中，模型输出还可以是一个离散值，例如图片类别。对于这样的分类问题，我们可以使用分类模型，例如softmax回归。和线性回归不同，softmax回归的输出单元从一个变成了多个。本节以softmax回归模型为例，介绍神经网络中的分类模型。Softmax回归是一个单层神经网络。 diff --git a/chapter_deep-learning-basics/softmax-regression-gluon.md b/chapter_deep-learning-basics/softmax-regression-gluon.md index 119ebcc..4d461ef 100644 --- a/chapter_deep-learning-basics/softmax-regression-gluon.md +++ b/chapter_deep-learning-basics/softmax-regression-gluon.md @@ -1,4 +1,4 @@ -# Softmax回归——使用Gluon +# Softmax回归的Gluon实现 我们在[“线性回归——使用Gluon”](linear-regression-gluon.md)一节中已经了解了使用Gluon实现模型的便利。下面，让我们使用Gluon来实现一个Softmax回归模型。 diff --git a/chapter_deep-learning-basics/softmax-regression-scratch.md b/chapter_deep-learning-basics/softmax-regression-scratch.md index 3d9a849..86cae47 100644 --- a/chapter_deep-learning-basics/softmax-regression-scratch.md +++ b/chapter_deep-learning-basics/softmax-regression-scratch.md @@ -1,4 +1,4 @@ -# Softmax回归——从零开始 +# Softmax回归的从零开始实现 下面我们来动手实现Softmax回归。首先，导入实验所需的包或模块。 -- GitLab