The course could have been much shorter than it is because Andrew seems to be repeating his simple ideas a lot in the lectures. However, each simple advice seems important for practical purposes (I am willing to take Andrew's word for it).

It might be not that trivial. But some hand-on experience with some code might be good here as well. As many practice as possible would be beneficial to the learners, coupled with great explanations from Andrew that are already in place.

I think the quizzes should be considerably longer, to include more scenarios, and also have fewer questions that rest on technicalities (where some of the answers are almost correct except they misuse a term or give a wrong description).

While the simulation is unique and very useful feature of this specialization. I believe examples with data would add to the leaning experience by allowing a student to actually run the scenarios and experience the qualitative changes.

This course is at all an important part during the learning journey. The only reason why I not rate full 5 stars that the recommendation ramen little bit on high level and do not address typical frame conditions in real world projects.

Lots of practical stuff about training models. But you should try building a few models before doing the course. Otherwise, you may not fully appreciate how much time can be wasted unless you use Andrew's clear and logical approaches.

Lots of interesting and useful idea. Unfortunately the editing is poor and Professor Andrew Ng has gone a little bit repetitive in his talking in this course only. The two previous courses were great but this one is kind of dragging.

I am a big fan of the jupyter notebook assignments. I can understand that it could be hard to build python assignments for this course, but not impossible I think (maybe around error analysis, impact of artificial data synthesis...)

The content of the course lecture is great. The teaching is great. One problem is the quality of subtitles. The black background does not allow to see what is shown behind. It would be better if the background would be transparent.

Great material! Some of the videos went a bit long, and I think the point could have been made in much less time. However, overall this series has been great and I still got some very valuable info out of this course, so I'm happy.

The course was very well structured and Andrews explanations was wonderful as usual. The only thing I was missing was more practical hands-on in the form of a programming exercise or two to really demonstrates the different ideas.

Andrew really provided great content, but the edition of this course is not as good as the first two, sometimes you will hear some repetitive sentences or a long pause. Hope they can upgrade this part, all the others are terrific.

This course give a lot of useful practical advices on training a machine learning/deep learning models. However, some of the advices are rather subjective and experience based, and some of the homework answers are quite debatable.

this course provided very interesting insight into missing , incorrectly classified labels and also how existing models can influence the training of a new model which is on similar lines as the task the existing models performed

Great course for machine learning strategies in deep learning.

Several concepts which aren't discussed in other courses have mentioned .

Especially the new way of splitting the datasets, transfer learning, multitask learning etc.

Looking at practical examples is an enormous help and some concepts i learned here will undoubtedly be useful in the future, i just think there should be more of it. It's just really short both in duration as well as content

Good course but tests needs to be corrected by the teachers. The first one evaluated things that would only be seen until the second week. and the second one needs to be checked in the different answers/options they offers.

It was really good to know how to structure and tune the nn so as to achieve a better model. But, I felt that it had too much theory in it that is hard to remember every time a model is to be designed. Overall, it was good.

One of the quick and great course for individual and team for understanding how to handle and structure the machine learning project. how to improve accuracy and handle error such a wonderful course made by deeplearning.ai

For some assignment feedbacks, especially the 'false' ones there is no explanation. I understand this is to prevent 'memorize' answers, but this could be just nicer if there is an explanation for only the selected answer.

I rate the course high. Unfortunately many of questions (posed in the forum) have not been answered.

Her are some suggestions:

Have quiz after every lecture. That will firm up the concepts.

Give lesser help in assignments.

Gives a sense about improving the performance of Deep Neural Networks, with error/bias/variance/data mismatch analysis. However, there is a lack of hands-on exercises, not having a programming assignment, only quizzes.

Lots of useful tips and tricks in this course. I feel that the videos could have been a bit shorter, and it would have been nice to have some programming assignments. Overall the course was extremely useful, however.

A pleasant diversion into practical considerations of project design. However the lack of programming assignments and the somewhat vague and fiddly quizzes make this a less satisfying course than it could have been.

This was a good overview of the concepts I have already learned. It was a good refresher on progress and changes in training best practices. There are a few flawed questions in both quizzes that need to be fixed.