学生对 deeplearning.ai 提供的 神经网络与深度学习 的评价和反馈

This is a nice follow-up to Andrew Ng's Stanford ML course. This one digs deeper into neural networks specifically, so if that's what you're interested in, this is a great course to take.

Note that the Stanford course used Octave and this course uses Python and NumPy (in Jupyter notebooks), so this is also nice because it gets you accustomed to using technologies that are more similar to what real ML practitioners are using. This course does still have you implement things by hand with NumPy and does not delve into higher-level frameworks like TensorFlow. For that, you will have to wait for the next course in the Deep Learning Specialization.

This course gave me an excellent overview of Neural Network, from the metaphor idea to math and implementation in Python. At least for me, the best way to study was a mix of pencil & paper (test and prove all equations) and reproduce the codes in the Coursera platform and Google Colab. The practice assignments are very related to theory lessons (equations using the same notation) that help the understanding. Only one note about the issues in notebooks, the Numpy version adopted is not the most recent, thus it is necessary to change some little things in order to reproduce the practice assignments on Google Colab (but this is not a problem).

This course is amazingly paced and also strikes a very good balance between required knowledge and depth of the topics covered. I cannot imagine how to improve this course except by asking for "more of the same". I had enough background in math and computer programming and I read already some articles and tutorials on Neural Networks. But only after this course I grasped the concept a little better. Andrew Ng is a very good educator: always ready to trade one pound of mathematical rigor for an ounce of intution. And I believe this is the only way to provide good contents here on Coursera. I strongly encourage everyone to take this course.

An excellent course to dive theoretically into basics of deep learning and also develop good intuitions about neural networks. Intricate details of linear algebra and the mathematical equations involved are neatly presented throughout the course. The programming assignments are meticulously developed to provide a very comfortable interface cum understanding of the problem, enabling the course learner to implement deep learning models on interesting set of classification problems. Adding couple more such problems (where one would use the already developed models) can make the practical learning experience even better. Enjoyed the course!

Very practical approach, full of code examples. It teaches you how to implement the NN with multiple layers from scratch in incremental steps. From the easiest approach (with single layer) to multiple layers. The code uses mainly simple code structures (i.e. loops, dictionaries, lists, vectorized operations and functions), so you do not need knowledge in OOP. Although I think some concepts if explained in OOP framework would be easier to grasp. But this is my subjective opinion. The course material is very well explained. If you want to learn and understand the way neural networks from inside out this course is definitely worth taking.

I've tried other introductions to deep learning courses, and they seem to focus too much on math or too much on coding - assuming the student is coming from one discipline or the other. This course nicely addresses both the math behind the algorithms, and the code required to implement it, without delving too deeply into either and focusing on the core of DL. This course uses Python and the libraries commonly seen in Kaggle kernels, and includes interviews with some of the most prominent names in AI, making it very relevant in 2018. I took the machine learning course from the same instructor and enjoy the delivery and organization.

As always, Andrew Ng's explanations help to grasp the material quickly and effectively. The programming exercises are interesting, yet not too challenging.

The course is, however, a bit light on the theoretical side. So if you are a practitioner looking for "hands-on" experience to get started with deep learning, by all means, this is your course.

If on the other hand, you are looking to understand the theory behind some of the concepts (i.e., you are not to afraid of a bit of math and would like to, e.g., see the derivation of the backpropagation algorithm), this course alone might not satisfy you. But it's a good start nevertheless.

This course is excellent and it is a great introduction to deep learning. Every week you learn new techniques and at the end of the course you are able to build a real deep learning application. If you have a solid math background you will gain a better intuition about the details of the algorithms. Finally, Professor Andrew Ng explains the content clearly and shares several best practices as well as useful advices that will make your learning experience very rich. I've loved the heroes of Deep Learning interviews and it is a great plus. I definitely recommend this course and I can’t wait to start the next one of the specialization.

Amazing course! I have tried learning concepts of neural networks by creating a syllabus for myself which consisted of different resources over the net. However at some point or another I would always reach a big obstacle which would prove to be extremely difficult to surmount and I would always inevitably give up. This course is structured in such a way that respects the current level of the learner and guides the learner through all the concepts without it being impossibly difficult or too easy. This course is only the beginning and I would gladly continue pursuing the other courses to strengthen my deep learning foundations!

As a teacher myself, I am impressed by how well organized is the course and how well they designed the assignments. Think they are introducing new knowledge to laypeople and they do it very well. However, I would like to get to know more about why neural networks work? In the content, there is a lot of the basis but you do not get to know where the magic comes from? I also love the interviews with the heroes of machine learning. That is something that really takes this course out of a purely instrumental one. Thanks a lot. The course fulfills my purpose of getting to know deep learning and keep me motivated to keep learning.

This is the best online course I have even attended. The instructor can explain advanced technical concepts in an absolutely easy and intuitive way. The instructor also can summaries the most important core concepts using graphs and diagrams which let students understand the core ideas on-the-spot and have prolonged impression. The lab exercises are organised and have a lot of guidance which is very very useful. The guidance is even until code-level which is very helpful in guiding students to produce efficient code. The lab exercises are also integrated with the real-world context that mimics the practices in the industry.

This course was awsome, brilliant , hatsoff to all the instructors , very well organized course special for those who are new to deep learning like me.When I start this course I am not confident very much , as I do progress I became confident specially after implementing the concepts to assignments , my math's knowledge about derivatives also helps me a lot . Suggestion - a video about matrix dimension in week 4 , I think this video should be in week 2 or week 3 because I have a problem how the dimension are set to W, b, Z .one small video about it like in week 4 but for 1 hidden or 2 hidden layer network. Think about it

Really great intro to Neural Networks. Andrew Ng, ( who is a Deep Learning professor and Standford, and co-founder of Coursera) walks you through all of the basic theory of Neural Networks then each weekly assignment you get a framework and then write the code from scratch in Python code from a single level Neural Network all the way up to a deep level network. If you're like me and aren't happy just calling the API without having some understanding of what's actually happening on the inside, this is the AI course for you. And with AI projected to increase the global GDP 14% (13.7T) by 2030 it really is the next step.

Chose this course to break into ML, coming from a background in quantitative MRI analysis. The videos are very well structured, and I've found the labs well-designed for cementing understanding of algorithmic implementation. I also appreciate how the course provides concision without skimping on the maths, where required. Having completed this first course, I feel like I have a confident grasp of some neural theoretical fundamentals, and my ability to implement algorithms leaps forwards with each lab I complete. I am aiming to complete the remaining 4 courses in this specialisation, as well as the Coursera ML course.

Andrew presented the course material in a very structured and systematic manner. The material is definitely a bit heavy, but Andrew does a great job in motivating the solution strategies. The systematic breakup of the backprop system of equations is probably the toughest part of the course, but that too was well-guided and the intuition was explained very well. I had some basic understanding of neural nets coming into this course, but I learnt a lot -especially the implementation aspect. Overall -this icourse had a perfect blend of theory and implementation for me to feel like I can now implement my own Neural Nets!

Andrew does a fantastic job of making this material accessible. This course is a great introduction to deep learning and won't overwhelm you with the details of the underlying mathematics. If you understand some fairly basic linear algebra and know how to take derivatives you'll be fine. The lectures are incredibly clear, and this is one of the best Coursera classes I've taken. The only critique I have is that the homework could be a little bit more challenging - or (if that would undermine the introductory nature of the class) there could be additional optional problems that push students a little bit harder.

A review from a business student with some programming and statistic foundation.

The programming assignments are great, guiding you to build part by part of the model.

Whenever you feel unsure what to do, make sure you read the instruction carefully, as clues/hints are often in there.

It's feels so awesome that I could finally construct deep neural network by myself instead of using packages that I have "some kind of" idea what's happening behind the scene.

Thank you Andrew! Your courses really inspire me, and when I become a master some day I will share my knowledge and experiences to inspire younger generations!

Good introduction to deep learning.

As Andrew said, the hard part is correctly deriving the matrix derivatives. not a easy job because it needs many prerequisites.

When I did programming assignments, I was not familiar with python syntax and had to google sometimes. I hope that Andrew includes more useful tricks/techniques that are commonly used in real life programming. I am not a programming beginner. But after years being in school, I am kind of less sharp than before.

The deadline gave me some pressure too. I am having more responsibilities as my family size getting bigger.

Thanks Andrew's team again.

This course is really an essential first step to AI. Using Logistic Regression to kickstart is a great way to demystify Deep Neural Network. One of my greatest weaknesses in learning Deep Neural Network was keeping track of correct dimensions in matrices. This course has a special topic on that, very thoughtful indeed. Having taken Geoffrey Hinton's Neural Networks for Machine Learning, I still consider the programming assignments to be very challenging but there are plenty of materials that helped me getting through it. All in all, this is a timely, thoughtful and extremely effective Deep Learning course.

Andrew NG is the most humble and talented teacher I ever came across. This course is paced right for beginners like me, prior to this course I had taken his Machine Learning course. I had basic ideas of logistic regression and Neural Network before. But this course enhanced my learning and also Python is a big help. (though sometimes i have to look for documentation even for most simple things, like getting a random array of certain dimension, but it aint a big deal). The core of this course is the understanding of forward and backward propagation. Which Andrew did with great details and make it simplified.

The instructor is exceptionally knowledgeable in the field. The way he explained the concepts was superb. All the difficult concepts were made exceptionally easy. I thoroughly enjoyed the course and learned many things throughout. Also the assignments on the way were very helpful for a hands-on session. They were a means to apply the concepts we learned and build confidence. Looking forward to learn more concepts in next course of the series. And I'd like to say thank you to Andrew Ng and to all his colleagues who are directly or indirectly involved in the creation of this beautiful course. Thank you all!

Such a great course! I had some basic khnowledge but without a systematic view. This course totally made me more familiar with the foundation and theory of deep learning. I am so grateful.

One thing I think can be improved is the tips and hints in the programming assignments. It appears the instruction are very detailed and I think the team can consider making a harder version of the programming assignments for those experienced students.

Moreover, the Jupyter Notebook is fantastic, but it can be hard to navigate through the window, so maybe an outline view (with all the function names) would be helpful!

Amazing course. Andrew has really streamlined the concepts, made the course easy to follow and at the same time leaves room for further analysis and curiosity. It is so well structured that can transfer complex concepts easily to you and therefore maintain the excitement in the student to keep on going at his/her own speed. What I loved most about the course was the fact that for some reason it seems like Andrew knows where to give you further explanation about what just happened or where you might get stuck in the code and in the lecture. Thank you Andrew. Such an amazing experience and great structure.

To start speaking, this is a really good course.

It guides through the basics of how to build a neural network instead of just importing from sklearn library. It helps one understand what is happening behind the scenes when one imports models from libraries like sklearn, PyTorch, etc. This course has taught how efficiently one can decrease computation time using vectorization as it made programming that enjoyable and exciting, it also reduced the time taken to complete the program exponentially.

I would recommend everyone who wants to learn how a machine learning model works and also build their own model

I found this course to be extremely good. It covers nicely theory, implementation and application of Neural Networks and Deep Learning. Prof. Andrew Ng through his video lectures makes it fun and easy to learn this subject with the right emphasis on key points. The quiz's and program assignments are really good, reinforcing the concepts. In addition I found the Hero's of Deep learning conversation videos towards the end of each week, informative and thought provoking. This is my second course after taking Machine learning on Coursera. I am enjoying learning on Coursera. Thank you Prof Andrew and Team.