great

The course started well but went downhill in week 3. The videos, actually get shorter and the treatment provided to the material related to Wasserstein distance, 1-L Continuity, interpolation and other crucial topics is just superficial. There are not adequate number of quizzes to test yourself. There is insufficient mathematical rigour. And it is too easy to pass the graded assignments without actually understanding the material. The forums are somewhat dead and you need to go to the Slack rooms to ask questions. On slack, it is a case of people linking to other papers rather than providing simple, direct answers. Nobody knows anything for sure. Overall, there is a take-it or leave-it attitude in this course and it is a far cry from Andrew's original ML Course which made Coursera such an attractive learning destination. I do hope the course is improved over time by adding more quizzes, delving deeper into topics (it's okay to have long videos where the instructor explains things slowly) and providing a more mathematically satisfying experience where the foundations are made stronger.

On the positive aspects - the notebooks provided are an excellent starting point to begin your own explorations. And the material is cutting edge.

El aprendizaje no ocurre desde lo más simple a lo más complejo. Simplemente se proponen videos uno después de otro sin evaluaciones formativas que efectivamente fijen el aprendizaje y sean consecuentes con la evaluación sumativa. No hay relación entre ambos tipos se evaluación ni en la dificultad que estas presentan.

En la primera tarea se evalúan aspectos que son explicados recién en la segunda unidad, ver los videos nuevamente no ayuda a entender el código que se presenta en la tarea, además se usan funciones para las que no se explica en detalle su funcionamiento.

Las lecturas paper, simplementes están linkeados en el curso, no se realiza ningún análisis de los mismos y no se elabora ninguna "bajada" del mismo que permita facilitar su comprensión. De esta manera es difícil que aporten algo al aprendizaje.

Pros:

*A lot of references to important articles.

*A lot of code in the notebooks that might be useful in the future.

Cons:

*The videos lectures are not comprehensive. This is sort of "self learning" course where one should read the articles on its own in order to really understand things. This is not what I am expecting from an on-line course (and it is also not like what I got used to from the DL specialization).

*Where are the pttx? I want to print them and write some comments

*The "labs" are basically a summary of some concept. There is no added value in writing them in notebook format since the code block is just "lets load this and this, and run".

1. Sacrifice width for depth  - There are so many additional optional readings (like in week3) where you have simply suggested papers to read. In my opinion, this could be replaced with in-depth discussions. As example is to discuss about the actual training in the assignment notebooks.

2. As ML engineers and practitioners, we are interested in knowing what solutions to adopt when problems occur in practice during the training. How to diagnose a traning failure and what are the remedies for it.

The examples and content in the course are excellent, but the assignments leave a lot to be desired. I spent more time debugging python than I spent debugging GANs during the assignment. This is not due to a lack of python knowledge IMO, but due awkward assignment structures that provide only cryptic feedback when inputs are not exactly as expected. To a colleague taking the course I might recommend them watch the excellent videos and play with the notebook examples, but avoid the assignments.

Concepts are explained wella nd clearly, which I appreciated, but to get a real understanding of things, a ton more of coding would be needed. In the assignment every thing is already cooked up, and you literally need about less than 20 lines of code to complete. This is a really weak point of thecourse in my opinion, since you end up NOT being able to implement things you saw in the lectures and in the related assignments

I liked the way instructor gives lectures but one problem is unfortunately she is not explaining things widely . Another problem is programming exercises. The problem is that you cannot print your code without writing it in true way which makes really hard to debug your code. Assertion codes are not informative. And there is not a expected result info as in other courses.

Course is pretty comprehensive considering that it is the first of a trilogy covering GANs. However, I would have liked to see more math. Reading the papers did offer the depth I wanted, but I would have liked to see the lectures break down the math in a more technical way rather than stop at shallow "intuitive" explanations and elementary school analogies

Under resourced course. Instructors do not reply to questions. If you have problems with the application, you are alone with very poor support and not clear reponsibilities on who can help you. Very disappointing and frustrating. Not practical information on how to deal with custom datasets provided. It is just a tutorial with an "easy" example.

The course was a good high-level introduction to GANs. The lectures were clear and very well done, however the course lacked mathematical rigor. The in-lecture quizzes were trivial, and so were the programming assignments. This course in isolation is not sufficient to get you started with GANs in the real world.

No se condice la pretendida profundidad de las explicaciones con las prácticas en código. Preferiría ir de a poco y más lentamente y dejar más claros los conceptos clave. La instructora es muy amable pero la velocidad del inglés es imposible de seguir para quienes no somos nativos.

The overview of several types of GAN with their potential issues that may arise, was good.

However, I would like to see the mentors more active in the discussion groups. I still have questions, that would have been answered quite easily by the mentors. That would have been great.

Thanks Coursera and DeepLearning.AI for providing this condition for all Enthusiasts.

This course would have been much better if the teacher had spoken a little slower, the scripts helped me a lot but there were some missing words in them.

The theoretical concepts are explained in a clear way, even if I would have liked a deeper dive into the math behind the loss functions of each model proposed, moreover the assignments were too guided imo.

Nice course overall!

The videos do not allow you to understand what is really happening at the code level, they explain the general theory, but it is not something that really allows you to create a project on your own from scratch.

Informative, but it feels like it didn't include explanations (or at least intuitions) required to fully grasp the concepts. For example, the necessity of 1L continuity and why does the enforcement work.

The videos teaches GAN, which is great, but the lab train for pytorch, which is great as well. But I wish the video and the lab works together so we can apply what we learn from the video into labs.

Very interesting and comprehensive courses, but more complex skill are in optional modules, perhaps these very interesting and intricate skills are for nex modules? Good course!

The programming assignments can be improved by designing it in such a way that most of the work should be done learner not by the course designer. I hope you change it in future.

Basically good, however the programming assignments are incredibly trivial compared to other machine learning courses I've taken on Coursera.

I would have liked to have more theoretical details on the mathematical point of view of the topics covered by this course

I don't know what the instructor is hurrying about. In terms of clarity, she is not even close to Andrew.

be unfamiliar with english and unlike Andrew use mathematical formula ， so i learn a little hard

Can be more detail. In week 3 and 4, there is not much information shared/taught.