使用 Python 进行深度学习路线图

Deep Learning Roadmap with Python
使用 Python 进行深度学习路线图

A guide to getting started with deep learning using the Python ecosystem
使用 Python 生态系统开始深度学习的指南

Freepik 弗里皮克

Machine learning is a good approach for small and medium datasets. But, when it comes to analyzing big data, this approach is unsuccessful. I learned machine learning techniques easily, but when performing deep learning projects, there are many challenges, especially for beginners.
机器学习是中小型数据集的好方法。但是，当涉及到分析大数据时，这种方法是不成功的。我很容易学会机器学习技术，但在执行深度学习项目时，存在许多挑战，尤其是对于初学者。

This blog will guide you to get started with deep learning. Here are the topics we’ll cover:
本博客将指导您开始使用深度学习。以下是我们将介绍的主题：

• What is deep learning? 什么是深度学习？
• Machine learning vs deep learning
  机器学习与深度学习
• Important deep learning algorithms
  重要的深度学习算法
• Deep learning applications
  深度学习应用
• Frameworks for deep learning
  深度学习框架
• Fields you need to know for deep learning
  深度学习需要了解的领域
• Awesome resources 很棒的资源

Let’s dive in! 让我们潜入！

What is Deep Learning? 什么是深度学习？

Deep learning is a machine learning technique and is used to find hidden patterns in big data. It is based on the artificial neural network (ANN) algorithm, which is a machine learning technique. ANNs consist of one hidden layer, while deep neural networks consist of more than one hidden layer.
深度学习是一种机器学习技术，用于发现大数据中隐藏的模式。它基于人工神经网络（ANN）算法，这是一种机器学习技术。人工神经网络由一个隐藏层组成，而深度神经网络由多个隐藏层组成。

Okay, we briefly talked about what deep learning is. Let’s move on to what is the difference between machine learning and deep learning.
好了，我们简单说了一下什么是深度学习。让我们继续讨论机器学习和深度学习之间的区别。

Deep Learning vs Machine Learning
深度学习与机器学习

Although deep learning is seen as a subfield of machine learning, there are some differences between them. Let’s take a look at these differences.
虽然深度学习被视为机器学习的一个子领域，但它们之间存在一些差异。让我们来看看这些差异。

Machine Learning vs Deep Learning机器学习与深度学习

Data Properties 数据属性

Machine learning is often used for structured data in tabular form, while deep learning is generally used for unstructured data such as images, videos, and text.
机器学习通常用于表格形式的结构化数据，而深度学习通常用于图像、视频和文本等非结构化数据。

Data Size 数据大小

Machine learning is used for small and medium datasets while deep learning is used for analyzing big data.
机器学习用于中小型数据集，而深度学习用于分析大数据。

Training Time 训练时间

While machine learning analysis takes a short time, deep learning analysis can sometimes take days or weeks.
虽然机器学习分析需要很短的时间，但深度学习分析有时可能需要几天或几周的时间。

Application Areas 应用领域

Machine learning is generally used for tasks such as classification, regression, and clustering, while deep learning is mostly used for tasks such as image, and NLP.
机器学习通常用于分类、回归和聚类等任务，而深度学习主要用于图像和 NLP 等任务。

Computing Power 算力

While machine learning projects can be performed with a simple computer, powerful computers are required for deep learning projects.
虽然机器学习项目可以用简单的计算机执行，但深度学习项目需要功能强大的计算机。

Cool, we have looked at the differences between machine learning and deep learning. Let’s move on to deep learning algorithms.
很酷，我们已经研究了机器学习和深度学习之间的区别。让我们继续讨论深度学习算法。

Deep Learning Algorithms 深度学习算法

Deep learning is more like art than science. You can obtain great models if you build the right architecture for your project. Let’s look at important algorithms used in deep learning.
深度学习更像是艺术而不是科学。如果您为项目构建正确的架构，则可以获得出色的模型。让我们看看深度学习中使用的重要算法。

Convolutional Neural Networks (CNNs)
卷积神经网络 （CNN）

CNN 美国有线电视新闻网

CNN architecture has revolutionized image processing. Thanks to this algorithm, computers gained human-level vision. This architecture uses filters and layers to extract patterns in images. Because of this way of working, this architecture is often used for the classification of images, recognition of objects, face recognition, and similar applications.
CNN 架构彻底改变了图像处理。由于这种算法，计算机获得了人类水平的视觉。此体系结构使用筛选器和图层来提取图像中的模式。由于这种工作方式，这种架构通常用于图像分类、物体识别、人脸识别和类似应用。

Recurrent Neural Networks (RNNs)
递归神经网络 （RNN）

CNNs performed well for image processing, but not for NLP tasks such as translation. The RNN architecture was developed for data where patterns change over time, such as NLP and time series. RNNs store historical data in memory and use this data to predict the next data.
CNN在图像处理方面表现良好，但在翻译等NLP任务中表现不佳。RNN 架构是为模式随时间变化的数据开发的，例如 NLP 和时间序列。RNN 将历史数据存储在内存中，并使用这些数据来预测下一个数据。

Generative Adversarial Networks (GANs)
生成对抗网络 （GAN）

Another groundbreaking deep learning algorithm is GANs which is used for increasing production, especially in fields such as art and video games. GANs are frequently used to generate data in areas such as image and NLP. For example, with GANs, you can generate photographs or realistic human speeches.
另一个突破性的深度学习算法是GAN，用于增加产量，特别是在艺术和视频游戏等领域。GAN 经常用于生成图像和 NLP 等领域的数据。例如，使用 GAN，您可以生成照片或逼真的人类语音。

Autoencoders 自动编码器

Today, there is a huge amount of data. It is very difficult to extract information from big data because it contains so many features. To reduce the dimension of big data, you can use autoencoders.
今天，有大量的数据。从大数据中提取信息非常困难，因为它包含如此多的特征。要减少大数据的维度，可以使用自动编码器。

Autoencoders learn a compressed representation of the data and enable processing on less dimensional data. For example, consider a thick book. You want to describe this book in just a few sentences. This is where autoencoders come in.
自动编码器学习数据的压缩表示，并能够对低维数据进行处理。例如，考虑一本厚厚的书。你想用几句话来描述这本书。这就是自动编码器的用武之地。

Transformers 变形金刚

Transformers is a deep learning architecture designed for NLP applications. This architecture reveals the meaning of the text by performing mathematical calculations and deals with different aspects of the text using many attention mechanisms. It worked great on architectural translation tasks.
Transformers 是一种专为 NLP 应用程序设计的深度学习架构。这种架构通过执行数学计算来揭示文本的含义，并使用许多注意力机制处理文本的不同方面。它在建筑翻译任务上效果很好。

Transformers 变形金刚

Transformers, when translating a sentence, first breaks the sentence into small parts and then performs many mathematical calculations to determine the meaning of each part and then puts these pieces together to form a complete sentence. State-of-the-art models such as GPT and BERT were developed using the Transformers architecture.
变形金刚在翻译一个句子时，首先将句子分成小部分，然后进行许多数学计算来确定每个部分的含义，然后将这些部分放在一起形成一个完整的句子。最先进的模型，如GPT和BERT是使用变压器架构开发的。

There are many more deep learning algorithms. In this section, we discussed the most important ones. Let’s look at the application areas of deep learning now.
还有更多的深度学习算法。在本节中，我们讨论了最重要的。现在让我们看看深度学习的应用领域。

Deep Learning Applications
深度学习应用

Deep learning is a technique that increases the ability of computers to learn and make decisions. This technique has many different applications. Let’s take a look at some applications where deep learning techniques are used.
深度学习是一种提高计算机学习和决策能力的技术。这种技术有许多不同的应用。让我们来看看一些使用深度学习技术的应用。

Image Processing 图像处理

Thanks to deep learning techniques, a computer can recognize objects in a picture and classify these objects correctly. For example, you can use deep learning in projects such as image classification, object detection, face recognition, and image restoration.
借助深度学习技术，计算机可以识别图片中的物体并正确分类这些物体。例如，您可以在图像分类、对象检测、人脸识别和图像恢复等项目中使用深度学习。

Natural Language Processing (NLP)
自然语言处理 （NLP）

Another area where deep learning is used a lot is NLP. For example, you can use deep learning in projects such as translation, text classification, text generation, sentiment analysis, and dialogue systems. The most popular application of deep learning in NLP recently is ChatGPT. This chatbot showed everyone what AI tools can do.
另一个大量使用深度学习的领域是NLP。例如，您可以在翻译、文本分类、文本生成、情感分析和对话系统等项目中使用深度学习。最近深度学习在NLP中最流行的应用是ChatGPT。这个聊天机器人向大家展示了人工智能工具可以做什么。

Speech Processing

There is no limit to what you can do with deep learning. Especially lately, deep learning techniques are used very often for speech processing. Speeches can be generated from your 3-second voice. You can also use deep learning in many speech processing projects such as speech recognition, speech synthesis, and music analysis.
深度学习可以做的事情没有限制。特别是最近，深度学习技术经常用于语音处理。演讲可以从您的 3 秒语音中生成。您还可以在许多语音处理项目中使用深度学习，例如语音识别、语音合成和音乐分析。

Health 健康

Deep learning also has many applications in the healthcare field. For example, you can use deep learning to diagnose diseases, produce medicines, and carry out health analytics.
深度学习在医疗保健领域也有许多应用。例如，您可以使用深度学习来诊断疾病、生产药物和执行健康分析。

Automation 自动化

Another use of deep learning is automation. With this technique, you can make projects in areas such as robotics and automatic control systems.
深度学习的另一个用途是自动化。使用这种技术，您可以在机器人和自动控制系统等领域进行项目。

Finance 金融

You can also use deep learning in finance. For example, you can use this technique for topics such as financial risk analysis, portfolio management, and credit risk prediction.
您还可以在金融中使用深度学习。例如，您可以将此技术用于金融风险分析、投资组合管理和信用风险预测等主题。

Agriculture 农业

Deep learning has many uses, but lastly, let’s look at agriculture. You can use deep learning for projects such as detecting plant diseases and classifying and predicting crops.
深度学习有很多用途，但最后，让我们看看农业。您可以将深度学习用于检测植物病害以及分类和预测作物等项目。

Nice, we covered some applications of deep learning. Let’s dive into the frameworks you can use to build deep learning projects.
很好，我们介绍了深度学习的一些应用。让我们深入了解可用于构建深度学习项目的框架。

Deep Learning Frameworks 深度学习框架

Deep learning analysis requires intense mathematical operations. It is very difficult to perform these operations manually. Fortunately, great libraries have been developed recently to carry out deep learning projects.
深度学习分析需要大量的数学运算。手动执行这些操作非常困难。幸运的是，最近开发了很棒的库来开展深度学习项目。

With these frameworks, you can train awesome deep learning models and monitor the performance of these models. Let’s take a look at these frameworks.
使用这些框架，您可以训练出色的深度学习模型并监控这些模型的性能。让我们来看看这些框架。

TensorFlow 张量流

The most used library is TensorFlow developed by Google. With TensorFlow, you can build end-to-end deep learning projects. This means that you can perform data preprocessing, modeling, and put the model into production with this framework. TensorFlow can be used with different programming languages such as Python, C++, and Java.
最常用的库是Google开发的TensorFlow。使用 TensorFlow，您可以构建端到端的深度学习项目。这意味着您可以使用此框架执行数据预处理、建模并将模型投入生产。TensorFlow可以与不同的编程语言一起使用，如Python，C++和Java。

Keras 硬

Another library we will go through is Keras. Keras was previously a library that worked on different frameworks. But it married TensorFlow in 2019. It is a high-level API in TensorFlow that even non-AI experts can easily carry out deep learning projects.
我们将通过的另一个库是 Keras。Keras以前是一个处理不同框架的库。但它在 2019 年与 TensorFlow 结婚。它是TensorFlow中的一个高级API，即使是非AI专家也可以轻松执行深度学习项目。

PyTorch PyTorch

Another open-source framework for deep learning is PyTorch. PyTorch is an open-source deep learning framework developed by Facebook. This framework quickly became popular because it was easy and flexible. In industry, TensorFlow is mostly used while PyTorch is often used for academic research.
另一个用于深度学习的开源框架是PyTorch。PyTorch是由Facebook开发的开源深度学习框架。这个框架迅速流行起来，因为它简单灵活。在工业中，TensorFlow主要用于学术研究，而PyTorch通常用于学术研究。

Hugging Face 拥抱脸

The last framework we’re going to talk about is Hugging Face. Hugging Face is an open-source platform for NLP applications. Hugging Face aims to be the GitHub for data scientists. Within this framework, you can find both pre-trained models and many datasets.
我们要讨论的最后一个框架是拥抱脸。Hugging Face是一个用于NLP应用程序的开源平台。Hugging Face旨在成为数据科学家的GitHub。在此框架内，您可以找到预先训练的模型和许多数据集。

Train your model with the Transformers library in Hugging Face and put these models into production with the Gradio library in your own space. Voila! You can also use it in your computer vision projects with the recently developed the Diffusers library.
在Hugging Face中使用变形金刚库训练您的模型，并在您自己的空间中使用Gradio库将这些模型投入生产。瞧！您还可以通过最近开发的扩散器库在计算机视觉项目中使用它。

Cool, we’ve seen the most used frameworks for deep learning. Let’s take a brief look at which areas you need to know to perform deep learning projects.
很酷，我们已经看到了最常用的深度学习框架。让我们简要了解一下执行深度学习项目需要了解哪些领域。

What is Needed in This Field?
这个领域需要什么？

Deep learning is an interdisciplinary area consisting of several fields. Don’t worry, you just need to know these areas at a basic level. There are three main pillars of deep learning. These are applied mathematics, computer science, and domain-specific knowledge. Let’s go through these areas.
深度学习是一个由多个领域组成的跨学科领域。别担心，您只需要在基本层面上了解这些领域。深度学习有三个主要支柱。这些是应用数学、计算机科学和特定领域的知识。让我们来看看这些领域。

Applied Math 应用数学

Applied mathematics includes, in particular, linear algebra, probability, and statistics.
应用数学尤其包括线性代数、概率和统计学。

Linear algebra is a fundamental area, as deep learning models are based on matrix computations. Probability is used to understand probability distributions of data. Deep learning models predict probability distributions of data. So it is important to know probability theory for deep learning.
线性代数是一个基本领域，因为深度学习模型基于矩阵计算。概率用于理解数据的概率分布。深度学习模型预测数据的概率分布。因此，了解深度学习的概率论很重要。

Another important area for deep learning is statistics. Statistics is a field used to analyze, explore and summarize data. Techniques such as mean, standard deviation, data standardization, regularization, and metrics are frequently used in deep learning projects. So knowing fundamental statistical techniques helps you build better deep learning models.
深度学习的另一个重要领域是统计学。统计是用于分析、探索和汇总数据的字段。平均值、标准差、数据标准化、正则化和指标等技术经常用于深度学习项目。因此，了解基本的统计技术有助于您构建更好的深度学习模型。

Computer Science 计算机科学

Another pillar of deep learning is computer science. For deep learning, you need to learn a programming language. The most used programming language is Python. Since Python’s language is easy, many libraries are written in this language. You also need to know parallel programming techniques to carry out your deep learning analyses faster.
深度学习的另一个支柱是计算机科学。对于深度学习，你需要学习一门编程语言。最常用的编程语言是Python。由于Python的语言很简单，因此许多库都是用这种语言编写的。您还需要了解并行编程技术，以便更快地执行深度学习分析。

Domain Knowledge 领域知识

Another field you need to know for deep learning is domain knowledge. Big data is very important for deep learning. To build a good deep learning model, you need to understand and interpret data related to the field you are working on. Don’t worry, if you don’t gain insight into the dataset. You can get help from an expert in the field for this.
深度学习需要了解的另一个领域是领域知识。大数据对于深度学习非常重要。要构建一个好的深度学习模型，您需要理解和解释与您正在处理的领域相关的数据。如果您没有深入了解数据集，请不要担心。为此，您可以从该领域的专家那里获得帮助。

Nice, we briefly talked about the important areas for deep learning. Finally, let’s move on to resource recommendations for deep learning.
很好，我们简要地讨论了深度学习的重要领域。最后，让我们继续讨论深度学习的资源建议。

Recommended Resources for Deep Learning
推荐的深度学习资源

Everyone has a different way of learning a field. Some like to learn by watching videos, and some by reading. I love to learn both by watching and reading. There are many resources for deep learning, both video and written.
每个人都有不同的学习领域的方式。有些人喜欢通过观看视频来学习，有些人喜欢通过阅读来学习。我喜欢通过观看和阅读来学习。有许多用于深度学习的资源，包括视频和书面资源。

I’m trying to collect these resources on our GitHub page called resources for data science.
我正在尝试在我们的GitHub页面上收集这些资源，称为数据科学资源。

GitHub - TirendazAcademy/Awesome-Data-Science-Resources: This repo contains resources on data…GitHub - TirendazAcademy/Awesome-data-science-resources：此存储库包含有关数据的资源...

Are you interested in the world of data science? Do you want to stay up-to-date with the latest tools, techniques, and…您对数据科学世界感兴趣吗？您想及时了解最新的工具、技术和...

github.com

In this section, I’d like to talk about the books that I love to read.
在本节中，我想谈谈我喜欢读的书。

Hands-On Machine Learning with Scikit-Learn & TensorFlow
使用Scikit-Learn和TensorFlow进行动手机器学习

(Image by Author) （图片来源：作者）

The first book I would recommend for deep learning is Hands-On Machine Learning. When asked for a book recommendation for deep learning, this book usually comes to mind first. This book consists of two parts. In the first part, machine learning with Scikit-Learn, and in the second part, deep learning with TensorFlow and Keras. It is one of the rare books that combines theory and practice.
我推荐的第一本关于深度学习的书是《动手机器学习》。当被问及深度学习的书籍推荐时，通常会首先想到这本书。本书由两部分组成。在第一部分中，使用Scikit-Learn进行机器学习，在第二部分中，使用TensorFlow和Keras进行深度学习。是少有的理论与实践相结合的书籍之一。

Deep Learning with Python
使用 Python 进行深度学习

(Image by Author) （图片来源：作者）

Another book I would recommend is Deep Learning with Python. This book is truly a legendary book for deep learning. This book was written by the developer of Keras. This book is great to learn the basics of deep learning with examples.
我会推荐的另一本书是《使用Python进行深度学习》。这本书确实是一本关于深度学习的传奇书。这本书是由Keras的开发者写的。这本书非常适合通过示例学习深度学习的基础知识。

AI and Machine Learning for Coders
面向编码人员的 AI 和机器学习

(Image by Author) （图片来源：作者）

The third book I’d recommend is AI and Machine Learning for Coders. I think you’ll read this book without getting bored. The book begins by teaching direct AI techniques with TensorFlow. It teaches deep learning with step-by-step applications. If you’re new to areas like computer vision, NLP, and time series, this book is for you. This book also explains the TensorFlow ecosystem very well.
我推荐的第三本书是《程序员的人工智能和机器学习》。我想你会读这本书而不会感到无聊。本书首先教授使用TensorFlow的直接AI技术。它通过分步应用程序教授深度学习。如果你不熟悉计算机视觉、NLP和时间序列等领域，这本书适合你。这本书也很好地解释了TensorFlow生态系统。

The three books I mentioned above describe deep learning with TensorFlow and Keras. Let’s take a look at another awesome book written using PyTorch.
我上面提到的三本书描述了使用TensorFlow和Keras进行深度学习。让我们来看看另一本使用 PyTorch 编写的很棒的书。

Python Machine Learning with PyTorch and Scikit-Learn
Python Machine Learning with PyTorch 和 Scikit-Learn

(Image by Author) （图片来源：作者）

The book I’m talking about is Python Machine Learning. This book consists of two parts. In the first part, machine learning algorithms with Scikit-Learn are explained. The second part demonstrates deep learning techniques with PyTorch.
我说的书是《Python Machine Learning》。本书由两部分组成。在第一部分中，解释了使用Scikit-Learn的机器学习算法。第二部分演示了使用 PyTorch 的深度学习技术。

Wrap-up 总结

Deep learning is an exciting field. AI tools such as ChatGPT are developed with deep learning techniques. By learning these techniques, you can perform projects with recent AI trends in your work area.
深度学习是一个令人兴奋的领域。像ChatGPT这样的人工智能工具是用深度学习技术开发的。通过学习这些技术，您可以在工作区域执行具有最新 AI 趋势的项目。

In this blog, we explored what you need to know about deep learning in 7 steps. You learned the differences between deep learning and machine learning, important algorithms, application areas, frameworks, and fields you need to know for building deep learning models. You also saw excellent books to get started in deep learning.
在这篇博客中，我们通过 7 个步骤探讨了您需要了解的有关深度学习的知识。您了解了深度学习和机器学习之间的区别、重要算法、应用领域、框架以及构建深度学习模型需要了解的领域。你还看过了开始深度学习的优秀书籍。

The best way to learn is by doing. To master deep learning, you must practice a lot with real-world datasets.
最好的学习方法是实践。要掌握深度学习，您必须使用真实世界的数据集进行大量练习。

That’s it. Thanks for reading. Let’s connect YouTube | Medium | Twitter | Instagram.
就是这样。感谢您的阅读。让我们连接优酷 |中 |推特 |Instagram。