Provider: DeepLearning.AI
Certificate/specialization: Deep Learning
Content: Coding exercises (autograded, not peer reviewed), Quizzes, Final exam (quiz length)
Final course format: Same as other courses
Final course content vs. prior courses: Final content is covered by previous course content
Time taken: Two days
Familiarity with subject before certificate/specialization: Very familiar, I "do" statistics as a research scientist and can code at a basic level. Although I wasn't familiar with many of the specific ML algorithms, I have transferrable skills and can write code to the autograder's standard.
Pitfalls, high points, things others should know: There are five "courses", but one is only two quizzes long and is based on two case studies. If you do these on the first pass through they take maybe 20 minutes tops. I recommend doing that course last. That leaves only four actual courses. All of the concepts are explained very well and I found the coding exercises fairly straightforward since there is no guessing what they are looking for. They tell you what you need to do at each step, you just need to decide how to implement it. Anyone who has taken calculus-based statistics and can code at a basic level will likely find it quite straightforward. On the other hand, someone who has only done e.g. Sophia Stats may be more likely to struggle since knowing how to implement the step is based on your knowledge of stats, matrix algebra, and vector calculus (i.e. why you need tensor addition and cannot simply use + for two vectors, or when it says initialize to zero why you need a 0 for a single variable but [] for a variable that will hold a vector). This stuff can be learned, it just will likely take longer than a couple of days to finish. I recommend reading this primer on tensors and vector operations.
I really liked the integration of the jupyter notebook into the Coursera environment. I hated the external Skills Network environment for the IBM courses. You can run everything as you go and there are no surprises when you submit to the autograder (which you can do an infinite number of times). If your code is working inside the jupyter notebook, you will get 100 from the autograder. This was the best designed course I have seen on Coursera in terms of user-centric design. Because everything is hosted inside Coursera your notebooks are saved and linked to your account. You can close things and go back later and your progress so far will still be there.
1-10 Difficulty level: I'll go with a 6 for me personally since I didn't find it challenging (although it was very interesting) and was only effectively four courses long. It requires pre-requisite knowledge though. If you don't have the prerequisite knowledge I would give it a 7. It could even be an 8, but the class is so well designed it still holds your hand through the steps and the company maintains its own forum with help boards for their classes (they also have other non-ACE certificates). The difficulty is not with the content per see, in fact the ML/AI part is kind of "in the background", but with having the right prerequisite knowledge to know how to do the coding tasks. The theoretical part and the quizzes are fairly straightforward if you watch the videos (I watched 2 x speed). I would gladly take more courses from them if their other certificates got ACE Endorsements (cannot say the same about the IBM ones, v glad they are over).
Certificate/specialization: Deep Learning
Content: Coding exercises (autograded, not peer reviewed), Quizzes, Final exam (quiz length)
Final course format: Same as other courses
Final course content vs. prior courses: Final content is covered by previous course content
Time taken: Two days
Familiarity with subject before certificate/specialization: Very familiar, I "do" statistics as a research scientist and can code at a basic level. Although I wasn't familiar with many of the specific ML algorithms, I have transferrable skills and can write code to the autograder's standard.
Pitfalls, high points, things others should know: There are five "courses", but one is only two quizzes long and is based on two case studies. If you do these on the first pass through they take maybe 20 minutes tops. I recommend doing that course last. That leaves only four actual courses. All of the concepts are explained very well and I found the coding exercises fairly straightforward since there is no guessing what they are looking for. They tell you what you need to do at each step, you just need to decide how to implement it. Anyone who has taken calculus-based statistics and can code at a basic level will likely find it quite straightforward. On the other hand, someone who has only done e.g. Sophia Stats may be more likely to struggle since knowing how to implement the step is based on your knowledge of stats, matrix algebra, and vector calculus (i.e. why you need tensor addition and cannot simply use + for two vectors, or when it says initialize to zero why you need a 0 for a single variable but [] for a variable that will hold a vector). This stuff can be learned, it just will likely take longer than a couple of days to finish. I recommend reading this primer on tensors and vector operations.
I really liked the integration of the jupyter notebook into the Coursera environment. I hated the external Skills Network environment for the IBM courses. You can run everything as you go and there are no surprises when you submit to the autograder (which you can do an infinite number of times). If your code is working inside the jupyter notebook, you will get 100 from the autograder. This was the best designed course I have seen on Coursera in terms of user-centric design. Because everything is hosted inside Coursera your notebooks are saved and linked to your account. You can close things and go back later and your progress so far will still be there.
1-10 Difficulty level: I'll go with a 6 for me personally since I didn't find it challenging (although it was very interesting) and was only effectively four courses long. It requires pre-requisite knowledge though. If you don't have the prerequisite knowledge I would give it a 7. It could even be an 8, but the class is so well designed it still holds your hand through the steps and the company maintains its own forum with help boards for their classes (they also have other non-ACE certificates). The difficulty is not with the content per see, in fact the ML/AI part is kind of "in the background", but with having the right prerequisite knowledge to know how to do the coding tasks. The theoretical part and the quizzes are fairly straightforward if you watch the videos (I watched 2 x speed). I would gladly take more courses from them if their other certificates got ACE Endorsements (cannot say the same about the IBM ones, v glad they are over).
![[-]](https://www.degreeforum.net/mybb/images/collapse.png)
