It is a very good course. Andrew Ng explanations are very clear and easy to understand with a lot of good examples. On the other hand there are some confusions or errors in the backpropagation part of the programming assignment about LSTM.

I am rating this course 4 because It doesn't give any guidance about future career paths and next things to learn. The explanations are very good. I understood complex things like GRU, LSTM, Bidirectional RNN, attention model very well.

While a great introduction on RNNs, I felt there could be another week of lectures given the complexity of the algorithms being explained. Likewise, the programming exercises felt unpolished in some parts, like in the expected outputs.

Very good. To make it perfect, would have liked it for the Assignments to have less bugs (cf. LSTM backprop), and a longer course with extra weeks to present LSTM in the context of prediction (finance, weather, pattern recognition,...)

Great course! only negative is that problems would really hold your hand. I don't think there is any way I would pass a whiteboard test on any of this (then again a course to get me to that level would have to be double this length).

The course covers one of the most influential developments in deep learning in recent times, and does so in a thorough way, introducing majority of the relevant mathematics and methods necessary to build a variety of sequence models.

The course is really good if you want to get a good understanding on the basics of deep learning. It would have been great if the course had more hand's on assignments than fill in the blanks kind of assignments in ipython notebook.

I learnt a lot from this course and the whole specialization. I am grateful to the mentors and instructors. If coursera gives me opportunity I can also be mentor for the specialization to help the newcomers through the assignments.

During the first week, I think that a deeper explanation of the matrices' dimensions throughout the NNs should be given. Indeed, this would be helpful to understand some concepts.

Apart from that, it was an awesoome course, thanks!

The course was good except first week. I did not learn as much as I would like from the programming exercises of week 1. It could be nice to have 4 weeks instead of 3 for this course. Taking more time to cover the week 1 material.

Everything is perfect, the teaching is excellent, the only problem is the jupyter notebook, its sometimes difficult to debug issues and takes a lot of time and is kind of vague as well in terms of application of the lectures.

Thanks to Andrew Ng and his amazing team, this course challenged my fundamental knowledge of Tensorflow and ML algorithms in the most useful way. However, I expected more applying libraries to design the NLP recent products.

some topics not explained in detail. Not enough examples to understand some models completely. As an example, I didn't fully understand what are the parameters for the models, their shapes, and how they are used in the model

Great course! But I find the transformer assignment bit difficult. I think if the implementation would be covered in a 10-15 minutes video it would be a lot more easier to understand. Thank you for the great specialisation!

Overall it was pretty informational on introducing NLP to me. However, Keras was a little bit frustrating to learn at the beginning. I found out the forum was a very good resource to learn Keras syntax whenever I was stuck.

Great course! I do feel like I'm just scratching the surface of the types of applications that I can make. I think the coding segments still hold our hands a little too much, but you can't beat the clarity of the lectures.

The course lecture is grade but I hope the assignment is better in guiding structure, something the explanation is hard to follow, and the assignment should include the transfer learning instead of using the trained model.

Until chapter 4 it is straightforward to follow the lecture and the assignment, but RNN(especially NLP, and transformer chapter), it gets too difficult to follow what Andrew said and apply it to the following assignments

This was hard to keep up with, maybe too hard. The assignments' difficulty also was on a different level then the lectures maybe there more time should be put into the lecture videos as it was the case for DNN and RNN.

The videos and programming exercises were very interesting and insightful. My only complain is some of notebooks for exercises include errors and it was just a time-wasting task to find the "trick" to pass the grader.

The content is well captured and Andrew really helps build the required intuitions. But, the assignments are too guided. There is no room to struggle for solutions which often proves to be the main source of learning.

I think the transformer programming exercise of this fifth course is not as good as the others. The methods we must implement are not clearly explained and the research on these really took me a huge amount of time.

The exercises are too short and too basic for this course specifically. The task is a great learning experience but there is not much one would struggle with in terms of difficulty as there is too much spoon feeding.

Great video lectures, but practical assignments are a pain due to awful auto-grading system and programming expirience in Jupyter in general. Most of the time you'll be searching for an error that isn't really there.

Excellent course, excellent lecturer. Unfortunately some of the test data (week3/lab/trigger word detection/XY_dev/* CANNOT BE DOWNLOADED ... The programming lab sections are nice - sometime a bit too helped ... ;)