Introduction

Deep learning has become a prominent topic in the field of artificial intelligence. It is known for its prominence in areas such as "computer vision" and gaming (AlphaGo), even beyond human capabilities. In recent years, the focus on deep learning has also been rising, and here is a survey result to refer to. Here is a search trend chart for Google:

If you're interested in this topic, here's a good non-technical introduction. If you're interested in learning about recent trends, then here's a good summary.

In this article, our goal is to provide a learning path for all deep learning people, but also a path to exploration for those who want to learn further. If you're ready, then let's get started!

Prerequisites

It is recommended that before learning deep learning, you should understand some basics of machine learning. This article lists complete resources for learning machine learning. If you want a simple learning version. Then you can look at the following list:

• Fundamentals of Mathematics (especially Calculus, Probability and Linear Algebra)
• Python Basics
• Fundamentals of Statistics
• Fundamentals of Machine Learning

Suggested time: 2-6 months.

Step 1: Machine configuration

Before proceeding to the next step of learning, you should make sure that you have a hardware environment that supports your learning. It is generally recommended that you have at least the following hardware:

• A good enough GPU (4+ GB), preferably Nvidia
• A ok CPU (e.g. Intel Core i3, Intel Pentium may not be suitable)
• 4 GB RAM (depending on dataset size)

If you are not sure, then read this hardware guide.

Note: If you are a hardware player, then you probably already have the required hardware.

If you don't have the required specifications, you can rent a cloud platform to learn. such as Amazon Web Service (AWS). This is a good guide to deep learning with AWS.

Step 2: First test of deep learning

Now that you have an initial understanding of the field, you should take a closer look at deep learning. Depending on our preferences, we can choose from the following paths:

• Learn through blogs such as Fundamentals of Deep Learning and Hacker's guide to Neural Networks.
• Learn through videos, such as Deep Learning Simplified.
• Learn through books, such as Neural Networks and Deep Learning.

In addition to the above prior knowledge, you should also learn about some popular deep learning libraries and the languages in which they run. The following is a less complete list (you can get a more complete list by checking the wiki):

• Caffe
• DeepLearning4j
• Tensorflow
• Theano
• Torch

Some other well-known libraries: Mocha, neon, H2O, MXNet, Keras, Lasagne, Nolearn. For deep learning languages, check out this article. You can also check out lesson 12 of Stanford's CS231n for an overview of some in-depth learning libraries.

Suggested time: 1-3 weeks.

Step 3: Choose your own field

This is the most interesting part, deep learning has been applied in various fields and has achieved the most advanced research results. If you want to go deeper, then as a reader, the path you're best suited for is hands-on. This will give you a deeper understanding of what you know now.

Note: a blog and a hands-on project will be included in each of the following areas A required in-depth learning library and an auxiliary course. The first step is that you should learn the blog and then install the corresponding in-depth learning library. And then do the actual combat project. If you encounter any problems in the process, you can take auxiliary courses.

• Application of Deep Learning in Machine Vision
• Reference blog: DL for Computer Vision
• Actual combat project: Facial Keypoint Detection
• Deep Learning Library: Nolearn
• Recommended course: CS231n: Convolutional Neural Networks for Visual Recognition
• Application of Deep Learning in Natural language processing
• Reference blog: Deep Learning, NLP, and Representations
• Actual combat projects: Deep Learning for Chatbots, Part 1, Part2.
• Deep Learning Library: Tensorflow
• Recommended course: CS224d: Deep Learning for Natural Language Processing
• Application of Deep Learning in Pronunciation
• Reference blog: Deep Speech: Lessons from Deep Learning
• Actual combat project: Music Generation using Magenta (Tensorflow)
• Deep Learning Library: Magenta
• Recommended courses: Deep Learning (Spring 2016), CILVR Lab@NYU
• Application of Deep Learning in reinforcement Learning
• Refer to blogs and practical projects: Deep Reinforcement Learning: Pong from Pixels
• Deep learning library: Need openAI gym to test your model.
• Recommended course: CS294: Deep Reinforcement Learning

Suggested time: 1-2 months.

Step 4: Dig deep and learn deeply

By now you should have learned the basic deep learning algorithm! But the road ahead will be more difficult. Now, You can use this newly acquired skill as efficiently as possible. Here are some skills that you should do to hone your skills.

• Repeat the above steps and select a different area to try.
• The application of deep learning in other fields. For example: DL for trading,DL for optimizing energy efficiency.
• Use the mental skills you have learned to do something else, such as referring to this website .
• Take part in some competitions, such as: Kaggle.
• Follow some researchers,
• such as: RE.WORK DL Summit.

Suggested time: Unlimited.

Recommended resources:

• Complete Deep Learning book
• Stanford UFLDL Turorial
• Deep Learning in Neural Networks: An Overview
• Awesome Deep Learning github repository
• Yann LeCun's recommendations for Deep Learning self-study