It has been quite a good course to explain the tedious concepts of RNN.

The only reason for a 4 star is there is definitely quite some room to improve upon the content and quality to bring it up to the mark of the previous 4 courses. There are quite a few bugs in the assignments which need to be rectified for the benefit of everyone, hope that it shall be done soon!

Concur with other reviewers: this class was good, covering a lot of interesting material and with well-structured quizzes & assignments. But the lectures seemed to skip past the sorts of in-depth explanations I wanted, instead just getting to the end point of "this is what this looks like". So good, but not quite as good as previous courses in the specialization.

This course is great to get intuitive understanding of Word Embeddings, RNNs, LSTMs, GRUs and Attention Models.

You will have great explainer videos and some excellent programming exercises. The course does not make you an expert, but it does make you familiar with the above mentioned architectures, so you can independently code and try them on your own solutions.

Really cool applications to work on, but the videos got a little too much into specific applications that may not be relevant most of the time. It was all interesting, but it made this course a lot longer each week. I could have done without a lot of the specifics of certain applications, just because it will be hard to apply/remember the concepts anyways.

I like the course for the very deep understanding of sequence models. But I really didn't like some practical tasks (especially closer to the end of course) as they are quite difficult to debug. It takes much time and doesn't get the feeling you understand how to handle with problems. Would love to have some insights about improving models debugging skill.

All courses in this specialization are awesome. However, this last course feels a little rushed in comparison with the other 4 courses. While the first 3 courses raise your knowledge of ANN in preparation to the 4th one, it is a little more difficult to understand this 5th course. Likewise, completing the assignments is possible, but more frustrating.

Guys, just the truth: Content: Great. Mr. Ng: Great. Autograder: Complete and utter BS. If you rework the Infrastructure you will be big. If you further refuse to do so (literally thousands of complaints about the autograder in the forums -> nothing happens) you will not. Check out Scala courses approach with grading -> works like a charm. Cheers, JJ

The content is generally great and well worth it. I wish they would fix some of the errors, especially in ungraded exercises. You end up wasting a lot of time because of them. Perhaps the most frustrating aspect is navigating to the Jupyter notebooks. I wish the links to the notebooks were on the same pages as the Submission and Discussion links.

I learned a lot. I would give 5 starts but the jupyter notebooks were very very buggy. I spent half of my time on the homework going through the forums to find workarounds. It took away from learning the material efficiently.

Note that I think that this may be a temporary problem as a new platform was release Jan 2019. The content was terrific.

The course is very interesting and it gives an insight into recurrent neural networks (RNN). The practical exercises are interesting but I found them in a bit raw state compared to the previous courses of the Deep Learning Specialization. Nevertheless I would still highly recommend to follow this course. Thanks a lot to organizers.

Everything seems logical, except the programming assignments. Although I went through week 1 programming assignments only, I often had to face some problems with insufficient information. Lecture videos are easy to understand, but not all the details are explained. (This is the point where I need to find some information by hand.)

Week 1 to 3 were great, the last week (transformer networks) compressed too much information into too few videos. Also the programming assignment was the only one in the whole specialization that I could not enjoy. It was a pain, and I was more than happy to pass it with <100% (which was the only one where I did not reach 100%)

As a non-computer science background student, the course was overwhelming, I got lost in the equations most of the time. Maybe a lower level course could be considered before starting this one. Nevertheless, this was an informative course about sequence models. Lots of quizzes and programming assignments reinforced my learning.

Videos are great; but as usual TP are too guided (hence boring) and do not use today frameworks (Pytorch, tensorflow 2). TPs should either be completely coded by candidates (only introduction + resfresh on concepts + objectives) with evaluation on final accuracy/f1 score <or> they should be no TPs at all and more MCQ tests

Content her is great - the first week covers the basic RNN models in a very clear way and the assignments are interactive and interesting, building on the explanations in lectures. One downsides is that the production quality is poor and would benefit from some re-recording to remove bloopers and make it smoother to watch.

I wish it was a bit more interesting. It also kinda feels like Andrew has a bit of a problem himself in understanding the paradigms stated in this course, and that makes me feel somewhat confused as well. Would recommend for the math, the notations are weird and confusing sometimes but it is understandable for most parts.

This is a very interesting course with good explanations, which give a brief but sufficient introduction to sequential models like GRU and LSTM. One star is dropped because the CNN course (#4) is still better than this one in terms of explanations, while course #2 is better in terms of relevant material and pace (to me).

Although I really really really love this series and although I always gave 5 stars, I think the quality of this last module is a lot less better than the previous ones. I think convolution was way more difficult but the explanation was awesome. Unfortunately, i think explanations in this module are a little sloppy.

As usual, Andrew Ng's stellar talent as an educator shines through. Unfortunately, some of the video editing is a little scrappy, and the assignments could use some more polish. Especially in areas where they catch quirks in the grader. Luckily the forum support is excellent. This course is definitely worth doing.

Overall nice course, learned a lot about NLP and Speech to text. Course is more oriented towards NLP applications, I was also hoping to learn more about time series analysis. Feel like the course could have been longer 4-5 weeks since RNN, LSTM and GRU is pretty long topic and 3 weeks seems to be too short for it.

Recomiendo ampliamente este curso, te proporciona un claro entendimiento de los modelos secuenciales y recurrentes. Es excelente, aunque a diferencia de otros cursos de esta especialización no explicaron en detalle algunos aspectos de las RNN, me hubiese gustado que profundizaran un poco más en backpropagation.

The RNN, LSTM< and GRU were very good. But the Week 3seemed a bit abstract. More could have been covered in Audio, Attention.

ALso the Jupyter Notebooks was frequently crashing, and it took lot of attempts to re-open the existing one. Lot of time wasted. Also it took long time to to submit and run the program

This course provided a nice high level overview of RNN models and associated Keras implementations. The tricks and tips given were a useful addition to my ML arsenal. The only thing that I feel discredits this course is that the programming assignments are easy to gloss over and pass without much engagement.

Great introduction to RNNs and how to implement them in keras. I suspect it is a relatively new course as there are still typos and a few errors in the assignments (otherwise I would have given 5 stars) but the forums help you to find your way around them and hopefully in future versions they will be fixed.

Excellent lectures, some part was difficult and it took time for me to imagine the content of each parameter (e.g. when we talk about X, or a, or Waa what is the size of them and what do they represent). But in the exercises, it became more understandable. Exercises need previous knowledge of Keras and OOP.