吴恩达 神经网络和深度学习 观看笔记（1）

Al is the new Electricity

Electricity had once transformedcountless industries: transportation,manufacturing, healthcare，communications, and more.Al will now bring about an equallybig transformation.

---

1、What is neural network?

lt is a powerful learning algorithm inspired by how the brain works.
这是一种强大的学习算法，受到大脑工作原理的启发。

Example 1：single neural network

一个现实的例子，通过现有的房屋面积和房价去生成function，给出一个size of house，可以自动生成一个price。

Given data about the size of houses on the real estate market and you want to fit a function that willpredict their price. lt is a linear regression problem because the price as a function of size is a continuousoutput.
给出关于房地产市场上房屋大小的数据，你想要拟合一个函数来预测它们的价格。这是一个线性回归问题，因为价格作为大小的函数是一个连续产出。
We know the prices can never be negative so we are creating a function called Rectified Linear Unit(ReLu)which starts at zero.The input is the size of the house (x)The output is the price (y).
我们知道价格不可能是负的，所以我们正在创建一个叫做修正线性单元(Relu)的函数，它从零开始。输入是房子的大小(X)，输出是价格(Y)。

图中的小圆圈就是神经元，“神经元”通过训练，实现了函数Relu(蓝线)。输入一个x，通过神经元的计算，得出一个y。

Example 2 ：Multiple neural network

The price of a house can be affected by other features such as size, number of bedrooms, zip code andwealth. The role of the neural network is to predicted the price and it will automatically generate thehidden units. We only need to give the inputs x and the output y.
房价可能会受到其他因素的影响，如大小、卧室数量、邮政编码和财富。神经网络的作用是预测价格，并自动生成隐藏单元。我们只需要给出输入x和输出y。

2、Supervised learning for Neural Network

In supervised learning, we are given a data set and already know what our correct output should look like,having the idea that there is a relationship between the input and the output.
在监督学习中，我们会得到一个数据集，并且已经知道我们的正确输出应该是什么样子的，因为我们认为输入和输出之间是有关系的。

Supervised learning problems are categorized into “regression” and “classification” problems. In aregression problem, we are trying to predict results within a continuous output, meaning that we aretrying to map input variables to some continuous function. In a classification problem, we are insteadtrying to predict results in a discrete output. In other words, we are trying to map input variables intodiscrete categories.
监督学习问题分为回归问题和分类问题。在回归问题中，我们试图在连续输出中预测结果，这意味着我们试图将输入变量映射到某个连续函数。在分类问题中，我们不是在预测离散输出的结果。换句话说，我们试图将输入变量映射到离散类别中。

以下是一些监督学习的例子：

There are different types of neural network, for example Convolution Neural Network (CNN) used oftenfor image application and Recurrent Neural Network (RNN) used for one-dimensional sequence datasuch as translating English to Chinses or a temporal component such as text transcript. As for theautonomous driving, it is a hybrid neural network architecture.
有多种类型的神经网络，如常用于图像应用的卷积神经网络(CNN)和用于一维序列数据的递归神经网络(RNN)，用于将英语翻译成中文或文本文本等时态成分。对于自主驾驶，它是一种混合神经网络结构。

Structured data refers to things that has a defined meaning such as price, age whereas unstructureddata refers to thing like pixel, raw audio,text.
结构化数据是指具有定义意义的事物，如价格、年龄，而非结构化数据是指像素、原始音频、文本等。

结构化数据一般在数据库中存在，而非结构化数据一般更贴近我们现实生活，容易被我们去捕获和理解，但是对于计算机来说不是件容易的事。多亏了神经网络，多亏了深度学习，这些非结构化数据也能被机器去慢慢理解。

3、Why is deep learning taking off?

Deep learning is taking off due to a large amount of data available through the digitization of the society,faster computation and innovation in the development of neural network algorithm.
随着社会的数字化、计算速度的加快和神经网络算法的创新，深度学习正在兴起。

看上面的图，当数据量较小时，神经网络的学习效能差异很小，往往取决于使用组件的性能。当数据量增大时，它们之间的差异显现出来，不同性能和不同规模的神经网络算法体现了不同的训练结果。在超大量数据的训练后，性能也不再提升。大数据时代，各种终端都可以产生大量的数据，所以数据规模在一定程序上推动了神经网络和深度学习的发展。

要达到高水平的绩效，必须考虑两件事：
1.能够训练出足够大的神经网络
2.大量有标签的数据

The process of training a neural network is iterative.
神经网络的训练过程是迭代的。
当你有了新的idea，就马上去改代码，去做实验。很快就会有结果，快的十分钟，慢的一两天。

lt could take a good amount of time to train a neural network, which affects your productivity.Fastercomputation helps to iterate and improve new algorithm.
训练神经网络需要很长的时间，这会影响到你的效率，快速计算有助于迭代和改进新的算法。

附上课后选择题和编程作业