The course content is great but the ending lacks tensorflow implementation of regularisation, hyperparameter tuning, learning rate decay, etc. and aslo still not confident enough in those.

The final programming task might contain minor bug, passed all sub-sections, but the final one result didn't match with the provided results, better provide more info for easier debugging.

Good material. The exercises are a little bit easy. The worst part is that after the last assignment, the certificate is done immediately and you don't have a chance to correct any errors.

Excellent lectures, well prepared, very good examples, great teacher.

I would happily give it 5 stars, if not the constant issues with Coursera infrastructure, crashing notebooks/kernels.

The course is very interesting and fairly well laid out but some simple typos can cause some confusion and they have been there for a long time based on some info in the discussion forums

Nice course. Great introduction to hyper parameters in neural networks and also nice assignment on tensorflow. It would have been even better if they introduced tensorflow in more detail!

Good material and some very nice practical tips. A few typos here and there in the course material made it difficult at times to debug the code, which is the reason for docking one star.

Good covering of many implementation aspects of neural networks. I find the practical exercises to lean on the tedious side while not bringing a tremendous amount of learning themselves.

It was somewhat frustrating to spend so much time coding raw python, just to discover that TF can do all of that with one-liners. Nevertheless it was valuable to learn the nitty-gritty.

Nice explanation of Tensor flow. Hyperparameter tuning was explained in easy and robust way. Programming Assignment is tricky but forum comments helped a lot in resolving the problem.

Some extra information on various optimization algorithms will be good. Moreover, if there are links to some of the research papers and resources to dive into, it will help out a lot.

Excellent course, but I did not learn many new things (some just from a different angle). Maybe the course description should be updated to be more specific about the target audience.

Some very interesting material for beginners. At times it feels like concepts are being repeated over and over again, but there is enough new concepts to keep it worthwhile to repeat.

Not so much different with the materials in the Machine Learning course from Prof. Andrew Ng itself. If you don't have the time to finish the ML course, then you should take this one.

Great content but there are major problems with the final assignment. The one-hot encoding function tests force the output tensor dims to be different than what a later cell expects.

The mathematical explanations were very good. But the coding task is always left to do at once. If it can be set after the corresponding videos as a module it would be great I think.

It is an excellent course. The only weird thing it is that it uses Tensorflow 1 instead of 2. I get it why is it done, but perhaps it would have been more useful to keep using numpy.

Awesome material, and everything is well explained. I would've liked that the programming exercises were a little more challenging, though going through the code shines a nice light.

As always Andrew Ng's clearly explains all the concepts along with practical programs. I would strongly recommend doing this course for a good solid understanding of neural networks.

While video lectures are very well explain subject matter, practical assignments are pretty frustrating since most of the time you will be battling jupyter notebook and auto grader.

Very good course, although it'd be awesome if Andrew went over the backprop associated with Batch Normalization and perhaps a programming example of using Batch norm on my test set.

I loved the course but the tensorflow implementation was a bit weak, it passed in just 15 min video. I recommend this course but as i told before tensorflow migration is a problem.

The graded functions could be a bit more free form, forcing you to think more about it. I sometimes feel that I'm more solving the "template", than I am thinking about neural nets.

Personally, I would like to have more programming exercises on the things that are taught (Hyperparameter tuning, Regularization) in order to compare how different techniques work.

Just enough explanation of material to get started on using DNNs for my own tasks. Assignments are easy, though provide good explanation of what is occurring in each line of code.