Let’s cover the specific
module content that you’ll learn in
this first intro course or four modules; module 1,
ML in finance and building your first model
using Google Cloud as a crash course from zero
to building a regression model in
a hosted IPython Notebook
using Pandas DataFrames and scikit-learn.
This is the first hurdle with
any data scientist needs to jump,
which is getting data into
the Cloud and executing your first model.
Module 2, trading fundamentals.
Once you’ve tackled
the Google Cloud tech start is trying to give
you a deep dive into
financial trading concepts like quantum theory,
arbitrage and back testing.
I’ll be quizzing you at the end
so you want to pay attention.
Module 3, supervised learning with BigQuery ML and ARIMA.
Once you’ve reached this module,
you already have the basics of IPython Notebooks and
trading concepts so we’ll
expand on both theory and model tack.
On the theory side,
you’ll learn about ARIMA models,
which stands for auto
regressive integrated moving average.
They explain a given time series
based on its own past values.
You’ll also experiment building ML models
using just SQL or SQL with BigQuery ML,
which even supports deep neural networks as a model type.
This is useful if you want to quickly test if ML is
viable on your data without
creating a complex custom model.
Lastly, in module 4,
we’ll introduce you to neural networks as
a model type and learn why deep learning is so popular.
You’ll then apply these advanced model types using
TensorFlow in the other courses in this series.
A lot of learners ask what are
they expected to code as part of this course?
Well, there are three interactive
IPython notebook based labs
for the first intro course, they’re listed here.
Building a regression model in Python,
building a regression model and BigQuery ML using SQL,
building an ARIMA model in Python.
We’re using linear regression largely because one,
it’s an easy first model type to start with and two,
because we’re predicting for a numeric value,
in most cases here,
the stock price of a given company.
What about environments set up?
Do you need to install anything locally on your machine?
No, through our partnership with Google Cloud,
each of your labs,
we’ll be using a real Google Cloud account,
no additional cost and
compute resources to train and run your models.
If you’ve used Qwiklabs before,
you’ve seen the lab environment shown here.
You’ll be given a certain amount of time to use
the project resources before they expire,
and you follow along the lab instructions in
the Qwiklab while doing
the actual work in your provided account.
Since most of these labs are machine learning based,
you can expect to see IPython notebooks like
this one that you’ll be reading through and execute it.
If you’re unfamiliar with IPython Notebooks,
check the course resources for a quick primer or look at
the lab solution videos where
we go through all the lab steps.
One key point to note is that
all of these notebooks are available
publicly in our course repository
linked here in the course resources.
That means even after this course ends,
you can still bookmark and refer back to
the code for this course and future courses.
Just as it’s critical to go over
what exactly this course covers,
I find it equally important to talk
about what we’re not going to cover.
Machine Learning Trading and Google Cloud are
three very broad topics and
this series of courses navigates their intersection.
Nationally, we can’t be everything for all audiences,
so we’ll provide links for
newer data scientists to get up to
speed and advanced challenges
for gurus to show off their skills.
Specifically what you won’t see here are building
and implementing a production
grade profitable trading model.
Sorry, at the core of
highly profitable models is likely a series of very
private and expensive to collect training
data sets that feed into multiple models.
This course will teach you
the core trading concepts used in professional models,
but it’s up to you to beat everyone else in
the market to build
a better model of reality for a given stock.
Next, you won’t see
any TensorFlow code in this first course.
We will be covering that in the second and third courses.
On that same topic,
most of the advanced machine learning
topics like tuning models,
building deep neural networks,
long short-term memory networks,
and reinforcement learning models are
covered in courses number 2 and number 3.
If you already have a deep understanding of ML,
you may want to focus on learning
the new trading concepts in this course and
experiment in the labs on how you can beat
the benchmark given your ML knowledge.
Last week, if you’re new to machine learning in Python,
we won’t cover basic intro concepts
like what is a feature or
model input as there’s a wide universe of
great generic ML content out there online.
Check the course resources for examples there and then
come back here when you want to apply
that theory to trading.

让我们介绍一下您将在第一个简介课程或四个模块中学习的特定模块内容；

模块1，金融学中的ML，

以及使用Google Cloud作为速成课程构建您的第一个模型从零开始到使用Pandas DataFrames和scikit learn在托管的IPython笔记本中构建回归模型。这是任何数据科学家需要跨越的第一个障碍，即将数据导入云中并执行您的第一个模型。

模块2，交易基本原理。

一旦你解决了Google云计算技术的问题，它将试图让你深入研究金融交易的概念，比如量子理论、套利和回溯测试。最后我会问你一个问题，所以你要注意。

模块3，使用bigqueryml和ARIMA进行监督学习。

学完本模块后，您已经掌握了IPython笔记本电脑和交易概念的基础知识，因此我们将在理论和模型方面进行扩展。在理论方面，您将了解ARIMA模型，它代表自回归综合移动平均值。他们根据一个给定的时间序列的过去值来解释它。您还将尝试使用SQL或bigqueryml构建ML模型，bigqueryml甚至支持深层神经网络作为模型类型。如果您想在不创建复杂的定制模型的情况下快速测试ML在数据上是否可行，这是非常有用的。最后，在模块4中，我们将向您介绍神经网络作为一种模型类型，并了解为什么深度学习如此受欢迎。然后，您将在本系列的其他课程中使用TensorFlow应用这些高级模型类型。很多学习者都会问，作为本课程的一部分，他们希望编写什么代码？好吧，有三个互动的IPython笔记本电脑为基础的第一个介绍课程，他们在这里列出。用Python构建回归模型，使用SQL构建回归模型和bigqueryml，用Python构建ARIMA模型。我们之所以使用线性回归，主要是因为第一种模型很容易建立，第二种是因为我们预测的是一个数值，在大多数情况下，是给定公司的股价。环境设置如何？您需要在您的计算机上本地安装任何东西吗？不，通过我们与谷歌云（Google Cloud）的合作，我们将使用一个真正的Google云账户，不需要额外的成本和计算资源来训练和运行你的模型。如果您以前使用过qwiklab，那么您已经看到了这里显示的实验室环境。在项目资源过期之前，您将有一定的时间来使用它们，并且在您提供的帐户中执行实际工作时，您将遵循Qwiklab中的实验室说明。因为这些实验室大多是基于机器学习的，你可以期待看到像这样的IPython笔记本，你将阅读并执行它。如果您不熟悉IPython笔记本电脑，请查看课程资源以获取快速入门知识，或者查看实验室解决方案视频，其中我们介绍了所有的实验步骤。需要注意的一个关键点是，所有这些笔记本都可以在课程资源链接的课程库中公开使用。这意味着即使在本课程结束后，您仍然可以将本课程和未来课程的代码添加到书签中并进行参考。正如复习这门课所涵盖的内容非常重要，我发现讨论我们将不涉及的内容同样重要。机器学习交易和Google云是三个非常广泛的主题，本系列课程将引导它们的交叉点。在全国范围内，我们不可能成为所有受众的一切，因此我们将为新的数据科学家提供链接，以加快速度，并为大师们展示他们的技能提供高级挑战。具体来说，你在这里没有看到的是建立和实施一个生产级盈利交易模型。抱歉，高利润模型的核心很可能是一系列非常私人且昂贵的训练数据集，这些数据集被输入到多个模型中。本课程将教你专业模型中使用的核心交易概念，但要想为给定的股票建立一个更好的现实模型，则取决于你在市场上击败其他人。接下来，在第一节课中您将看不到任何TensorFlow代码。我们将在第二和第三节课上讲到这一点。在同一个主题中，大多数高级机器学习主题，如调整模型、构建深层神经网络、长短期记忆网络和强化学习模型都在第2和第3课程中涵盖。如果你已经对ML有了很深的了解，你可能会想在本课程中重点学习新的交易概念，并在实验室里进行实验，以了解你的ML知识如何超越基准。上周，如果您是Python中机器学习的新手，我们将不介绍基本的介绍概念，比如什么是特性或模型输入，因为在线上有大量优秀的通用ML内容。查看课程资源中的例子，然后当你想把这个理论应用到交易中时，再来这里。