I found this course absolutely excellent. The structure and approach are absolutely great, and I am very happy that you force students to understand the mathematical underpinnings of backpropagation instead of letting a DL framework do the heavy lifting for you. Engineers need to understand what they do deep down.

My only improvement suggestion would be to provide a more detailed explanation of why we do the matrix multiplication & transpose in the computation of dW and of dA[l-1]. It turns out that in the case of dA[l-1] the explanation goes to the heart of reverse-mode differentiation and how to avoid combinatorial explosion. Cfr's Colah's blog excellent paper on backpropagation for details.

I liked this course very much. I have done coding and trained models in Pytorch and didn't have strong grasp in the math's part i.e Gradient and derivates that is the why i have taken this course at the first place. Though the course doesn't contain everything but it has given me enough knowlegde to start with deep-learning.The quiz and programmning excersice are really good. I have to think enough at some part and have done mistakes many times but got my concepts cleared. And thanks to coursera team for approving my financical aid.And though review may be good or bad depending on the person but i have learnt what i want to learn and it is good enough rahter than youtube or online material.

I have done a bit of neural networks and machine learning already in uni, so this was a bit of revision for me. However, I still learned quite a few things. Andrew just explains it in a way that is very oriented to actually using the information in a project, or otherwise in a way that makes the info feel natural to digest. Things I remember struggling with in my uni course felt a lot easier after the course and it only took me 5 days to complete. That's not even a week of your time.

Long story short: even if you are not a complete beginner to the topic, I still recommend to take the course, as it is a quick refresher of all the relevant topics and takes relatively little time to complete.

Absolutely Fantastic. I thought the programming assignments were a little too easy, but that's probably because I am familiar with python programming. I must say that the structure of the code really helped me focus on the core algorithms and vectorization (using numpy methods), so, in retrospect, it is probably a good way to make the student focus on the core concepts. I wish, however, there were some (more) optional lectures on the math and some more detailed derivations and some "optional" practice problems on doing partial derivatives etc., just to cement some important concepts such as back propagation. Highly recommend this to students wanting to learn the basics of neural networks.

I gained so much valuable knowledge from this course through the weekly Assignments. Learning by doing is real. The lecturer prepared exciallent materials especially the programming assignments that give us clear guidelines step-by-step, making it easier to understand. He took the effort so much to prepare the course. Additionally, he explains all the complicated points clearly including the derivation of various equations that require calculus knowledge which I did not expect to see him explain. Hopefully, I will be able to find time to take the rest course in this series to find some ideas of neural networks and deep learning that can be implemented in my master's and Ph.D. research.

Awesome course. I have fairly little previous math experience though I have been working on some calculus and LA immediately before/while taking this, and all the topics were easy enough to understand how they are supposed to work. Much recommend.

One small thing I think could have helped a bit is the practical examples do a little bit too much hand holding. It makes it a little hard to know if you are actually grasping the knowledge or just able to tell what to do based on what information has already been made available from the templates and such. Had to step outside of this and try to do some of it on my own to see which pieces weren't fully making sense. But still, awesome course!

Great, just like the first Intro to Machine Learning course Professor Ng distributed. Same style with improvements made in course design. For example, notation is much more consistent this time around probably because it is a more focused course unlike the first one. I would say taking Intro to ML first would help as it is a perfect primer for this course. Also, I'm glad they've decided to use Python which is just much more general purpose than MatLab. I would also say a solid grasp of the language is needed as a lot of boiler-plate code is provided and understanding it could be difficult if not otherwise comfortable with Python. Looking forward to continuing the full specialization.

Amazing course! The creators are very good teachers. Materials have the right mixture of motivation, real world examples, theory and practice. I also like Andrews presentation style - one can really feel that he truly cares about the students being given good information and getting encouraged to learn. The assignments were also very well made - everything works, the code is good and there is so much help in the context/comments (eg. someone could even finish the labs without understanding the subject) - but this basically allows each student to choose how much effort he/she wants to put into the subject (also meaning how much knowledge she/he wants to absorb during this course...)

I found this course very helpful in furthering my understanding and clearing a few doubts that I had from the Machine Learning course. I seem to understand back propagation much better now.

This course also helped me give a structure to the steps involved in actually building a Neural Network... gives me more confidence.

My only issue was with the programming exercises. I felt they were very tightly structured, maybe because of the automated grading system. It was almost impossible to go wrong. More flexible and open exercises, I think, will help in learning the real intricacies of building a NN from scratch. Don't really know enough to comment on how this change can be incorporated

The course is well structured and the programming exercises are so detailed, I am going to refer to them in future while implementing neural networks. The best part about the course is, Andrew Ng actually taught the math behind the network. Rather than taking his students through a library function for neural networks in python, he taught his students how to code from scratch while also covering nuances such as suitable activation functions for different cases and ideal values for weights. The documentation for programming exercises is very detailed and must have taken plenty of time for those who worked on it. Recommend it for anyone wanting to start. Kudos to the instructors!

The programming assignments (things which the student had to fill in) seemed repetitive and very limited (e.g. mostly needed only mathematical formulas to be filled in using numpy). However, to keep the grading similar and also cater for less advanced users simultaneously, the assignments could be tiered by beginner-intermediate-advanced (by concealing more and more stuff) but still grade based on the current beginner level only. So, one could start with advanced and then reveal more to get to intermediate in case he is not able to solve etc .. May be optional bonus grades for solving it at advanced level etc.

Otherwise a good course overall to get intuition into deep learning.

I took the Machine Learning course last year, and my only complaint was that the neural networks part was a bit confusing. (Once the bias unit had to be in the weights matrix, once it had to be removed in order to perform some calculations.) The formulas also didn't work all the time, and you needed to figure out the proper shapes (using transpose operations) on your own.

In this course, the notation and explanation is much clearer. Keeping the bias unit out of the weights matrix in the first place makes everything easier. (Of course, the video quality is also better.)

Thanks Andrew; you really listened to your audience and figured out how to explain those concepts more clearly!

A very good entry-level course. Made me recall linear algebra and calculus but in a good sense :) Andrew does a great job in explaining the material and his "if you don't get it exactly - don't worry - there are plenty of deep learning practitioners who don't get it either and are still very successful" - is very reassuring. He is the type of the lecturer that explains to you the material in such a way that those who do not have enough background feel very comfortable as well as those who have enough background do not dose off either.

In general the course made me want to go on with the next courses in order to get more insights regarding what's going on inside deep learning.

After the "Machine Learning" course from Professor Ng I saw this specialization and gave it a try. He manages to keep a great quality of content through out the entire course and explains everything in a great understandable way. The interviews he added to this course were especially great to listen to and hear how some of the pioneers of artificial intelligence see their own field and hearing about their stories/origins motivates one to really dive into the deeper topics and also develop interest in a specific field of AI. I really recommend this specialization to everyone who wants to dive deeper into AI and even gain new motivation on how to approach such a complex topic.

This is a really useful course to learn the basics of Neural Networks and the intuation behind it!

I have been working with Neural Networks for some years now. Recently I started preparing for job interviews in the field of Machine Learning, so I wanted to refresh my knowledge about the "theory" behind deep learning. Surprisingly, this course made me even learn some concepts and ideas that I had never considered before!

Everything is explained in a very intuitive way. Normally, in this field, we tend to work systematically without even considering what we are doing or why we do it this way, This course helped me easily understand Neural Networks in a way I will not forget.

Oh My God, my first Coursera course that i have finished to the end.. Supper happy and supper excited till I go to the next one. It is so engaging that even working on a temperature above 30 in no AC room did not slow me down. I also started this course while i was officially in vacation since I could not wait till i get back from vacation. Anyways, Very good content, easy to follow and the fact that I had to implement all the theory right away was just super. I learned not only the power of NNs but also my favorite programming language Python. Any one with a motivation and interest in DP should take this course because it gives the foundation in the best way possible.

This tutorial was so elaborated. And in each week Andrew Ng tried to recap important findings from previous lessons which were helpful. Sometimes it looks fuzzy to recognize if the instructor is referring some notation as raw or vector form. But mostly it was nicely designed. I love the way programming exercise was designed. It can provide the basis to build a neural net from scratch. Considering all levels of users, he gently represented all the complex term like derivative in a simple way. Maybe for the future suggestion, Besides handwriting, if those calculations of those function can be displayed in animated design, then it's possible to make it simplified enough.

Andrew is an exceptional teacher. Thanks to him, I clearly understood the structure of a neural network and the functioning of the whole network starting from the single neuron.The mathematics behind a neural network, which until recently seemed very difficult to me, is now very clear.

This course is even better than the one on Machine Learning of Andrew Ng because here you can directly use Python with the Numpy library and all the part of the exercises and practice is, in my opinion, much better structured and clearer than the other course. I recommend it to anyone with an interest in Artificial Intelligence. I can't wait to continue my Deep Learning Specialization.

Andrew Ng is a fantastic intructor. I admire his teaching style. He pays so much attention to the fundamentals instead of rushing through the material, that I feel like I learned something that will actually stay with me. The homework codes are written beautifully. Introduction of broadcasting and vectorization was an eye opener - turns out I've been programming very inefficiently for years without knowing. This course on it's own is not enough for me to go and architect NNs on my own, but it definitely helps with general understanding of the process, I feel more confident now talking about it and reading papers. Will continue on to other courses in Specialization.

A must for anyone in deep learning research. This course aims to build the foundation of deep learning operations by not using the built-in functions but writing code yourself, which help tremendously later. It gives you the microscopic view of what calculations are carried at each neuron, layer, forward pass & backprop.

The interviews provide the right kind of motivation for aspiring researchers. They're like the cherry over the cake! The syllabus describes the course material but whats a plus in this course is Prof. Andrew Ng's tips when it comes to applying techniques and information about the latest (and probably near future) trends of the academia and industry.

After studying Lesson 1, I learned a lot and solved many problems I've been puzzled before. Andrew-NG's depth explanation and detailed writing move me deeply. Teacher's profound knowledge and responsible attitude is my learning example .The teacher can make the complex knowledge lively and interesting, but without losing its own contagion. After-class exercises design is also distinctive, providing great convenience for our beginners . After class, the active discussion and exchange provide a wide range of ideas and rich ways to me. Thank you, deep leaning team. we thank coursera for offering rich courses, thanks to Miss Wu's team for doing so excellent course.

A few months ago, I didn't know anything about machine learning. Machine learning context was a wonderland for me and I decided to dive in. I began with "The Machine Learning course by prof. Andrew Ng.

I thought about how can merge my knowledge of embedded systems with machine learning. I understood that the implementation of deep networks on customized hardware such as System-On-Chip is an open issue.

Now, I learning more about deep learning through DeepLearning.ai courses on Coursera.

I'd like to special thanks to Andrew Ng, Kian Katanforoosh, and Younes Bensouda Mourri for this curriculum.

I wish that finding a position for implementing deep learning on edge.

Considering how clear and thorough lectures by Andrew Ng were and overall how hard things were made simple in this specialization I can't give it anything but 5 stars. Thank you very much for your hard job on it!

However, I would prefer a bit harder and more theoretical course, personally. This one was adapted for a very broad range of listeners, which is a good thing generally. But it is absolutely not challenging to pass it: for instance, the programming excersices are great notebooks, but they mostly are already solved for you and you only need to fill the right lines into the right places. Only the last course on sequential models probably was a bit harder.

If one has already completed Andrew N.G.'s Machine Learning course that works on Octave & Matlab, then this course will be a piece of cake. However, the refresher here, is Python! And there are SO MANY things that course expects you to know - so much to learn! The Material is designed to NOT MAKE YOU UNCOMFORTABLE but if you really want to Learn Python, then you will have to take it as a challenge and learn pretty much everything that you see as unknown there.

Overall though, really excellent course material. Glad I picked up this course. And I think it is a good revision for one already versed with ML concepts that one can easily pick up this Specialization.

Very well designed and thought through course - Highly recommended for those who want to learn neural networks from scratch even extending it to deep learning.

This course will empower you to understand, create, and tune a neural network. Clearly describes about Parameters, Hyper-parameters tuning, Forward Propagation, Activation Functions, Backward Propagation, Updating Parameters and Predicting Labels.

On a side note :: Before this course I was only aware about analogy of human brain's neurons and neural network and after this course I am able to understand that no one knows (even neuro scientists) that what a single brain neuron does.

HaPpY Learning Guys !