Red Stone’s personal website: Redstonewill.com
Speaking of Andrew Ng, I believe you are familiar with him. As the big IP of artificial intelligence, Ng has been committed to the promotion and popularization of artificial intelligence, striving to make everyone feel the charm of artificial intelligence. Starting in August last year, Ng started the ai deep learning craze by running a five-course Coursera deep Learning course. Here is the website of Deeplearning. ai:
www.deeplearning.ai/
The quality of this series of special courses is very high. Red Stone has also learned these five courses completely, and systematically recorded all knowledge points in the form of notes, hoping to bring some help to you. A summary of all notes is as follows:
The final | Wu En da deeplearning. Ai all special course refined notes summary
It’s not over! The bull can’t stop. More recently, Ng launched CS230, a high-quality deep learning course at Stanford University. The homepage for this course is:
Web.stanford.edu/class/cs230…
The description of this course is:
Deep learning is one of the most sought-after skills in AI. We’ll help you learn deep learning. In this course, you will learn the basics of deep learning, understand how to build neural networks, and learn how to lead a successful machine learning project. You’ll learn about convolutional neural networks, RNNs, LSTM, Adam, Dropout, BatchNorm, Xavier/He initialization, and more. You’ll conduct case studies from healthcare, autonomous driving, sign language reading, music generation and natural language processing. Not only will you be able to grasp the theoretical knowledge, but you will also be able to see how it is applied to industry. All of these ideas can be implemented using Python and TensorFlow, and that’s what we’re going to teach you. By the end of this course, you will most likely find some creative ways to apply to your work. This course is taught in a flipped classroom format, where you can watch instructional videos at home, complete in-depth programming assignments and take online tests, then come to class for further discussions and complete projects. The course will end with an open final project, with some help from the teaching team.
This course should have the following basics:
-
Master basic computer science principles and skills, be able to write a reasonable and complex computer program.
-
Basic probability theory (CS 109 or STATS 116).
-
Basic linear algebra (Math 104, Math 113, OR CS 205).
This course is based on a combination of offline lectures and Coursera online learning. In short, this means watching instructional videos on Coursera, taking online tests and submitting programming assignments. Study other materials, discuss them, and finish an open project in the classroom. The Coursera videos in question are from the five courses in deeplearning.ai, which also feature online tests and programming assignments. Those of you who have done it before will have an easier time mastering it. Even if done, also suggest to do afresh, consolidate knowledge point, and the final project just is key.
CS230 consists of 5 courses. The course content is roughly the same as deeplearning.ai, but there are also changes and extensions:
1. Neural networks and deep learning
The first course focuses on the basics of neural networks and deep learning. It mainly introduces the intuitive concept of deep learning in class, and uses two modules to learn what a neural network is from scratch.
2. Improving deep neural networks: hyperparametric debugging, regularization and optimization
The second course mainly introduces the internal mathematical structure of deep learning model, and gradually transitions from shallow network to deep network to understand the significance of “depth”. Once you’ve mastered these concepts, you’ll have a basic idea of how to build a deep learning network from scratch.
Then there are optimization or tuning techniques for deep models, such as initialization, regularization, data set partitioning, Dropout, normalization, gradient checking, etc., and various classical learning rate decay methods, such as momentum algorithm, Adam, etc.
3. Build machine learning projects
The third course focuses on structured machine learning projects. The basic part covers hyperparameter tuning, batch optimization methods, and the application of deep learning frameworks such as TensorFlow and PyTorch. Then machine learning strategies, including vertical tuning, evaluation index setting, data set division, etc.
Convolutional neural network
This course mainly introduces convolutional neural network, which is mainly used to process spatial data, such as images and videos, so it is widely used in computer vision. There will be a midterm exam during this part of the course to help you review the material you have previously studied.
The basic part of CNN involves convolution operation, stride, pooling, etc., and then it further introduces several classic CNN architectures, such as Lenet-5, AlexNet, VGG, ResNet, Inception, etc. After that, several suggestions in the development process of CNN are given, including transfer learning and data enhancement. Finally, the current research status of CNN field is introduced.
5. Sequence model
This course focuses on sequence modeling. Sequence model is mainly used to process sequential data, such as music, speech, text and so on. Sequence model is mainly represented by recurrent neural network. This course will introduce the basic structure, type and calculation process of RNN, and analyze it with language modeling as a typical case. This was followed by some well-known variations of RNN, such as GRU, LSTM, bidirectional RNN, deep RNN, etc.
Final project of CS230:
One of the main goals of CS230 is to help you apply machine learning algorithms to real-world tasks, or to equip you for machine learning and AI research. The final project aims to get you started in these directions.
Fortunately, the project report and Poster presentation of CS230 have been released. Covers a variety of topics, such as music generation, mood detection, film emotion classification, cancer detection, etc. The final list of projects and the top three posters has been announced:
Of course, there are a lot of project submissions, so if you are interested, you can check them out:
Web.stanford.edu/class/cs230…
CS230 Course Materials:
The PDF files of all the course materials for this course, including the final project, have been collected for you. Access is very simple, as long as the wechat public number: AI Youdao (ID: Redstonewill) background reply: CS230 can be!
Conclusion:
For this course, I pay more attention to the final project, which can help you consolidate what you have learned and improve your practical operation ability, which is quite important. Finally, I hope we can consolidate the theoretical knowledge of deep learning and improve our ability to solve practical problems.
