Avis et commentaires pour d'étudiants pour Deep Learning and Reinforcement Learning par IBM

good course, PROS: very well presented, clear amd methodic. appropriate tasks. CON the name suggests that RL would be a significant topic, while in reality it appeared only in the end of the course and important subjects such as TDlearning are missing.

Reinforcement Learning part needs to be a separate course and more details in it

Well prepared, gives a good intro to multiple Deep Learning algorithms and good examples to cover the major topics. A few more practice labs on CNN and RNN would have been awesome!

Cons : The only difficulty I found was with the english accent of our dear trainer. Sometimes it was really very difficult to comprehend what was being said and one needed to rewind the video multiple times and read the subtitles. Other than that, nothing to complain.

Cheers!

The concepts were clearly explained in lectures. The assignments were very helpful to gain a practical insight of the skills learned in the course.

Extraordinary course, one of the best in coursera!, Reinforcement Learning and Autoencoders can have better examples.

E​xcellent from the theory and the practice! Great explainatory videos and detailed jupyter notebooks!

Hello, thank you again for the course. My congrats, once more, to the instructor on the videos!

The difficult terms are simplified enough for understanding and application in real life.

this course is awesome

Well balanced course

Excellent course

Very good. I learned a lot but the subject matter is quite extensive.

This course has a larger scope than the other ML certificate courses and is a little out of date. While it introduces RL, it does not discuss TD learning or Deep RL. RL seems "tacked on". Similarly, there is a brief introduction to Attention, but no substantial discussion of Transformer models (I suggest dropping LSTM and talking just about Transformers). Unlike the other courses, which introduced the concepts and also covered practical steps on using these methods, the DL/RL course is a little light on the practical side of DL. There is little discussion of why particular architectures are chosen for specific problems or how sensitive those architectures are to various hyperparameters. You will know what DL, CNN, RNN (and to a lesser extent, RL) are is when you finish this course, but there's a big gap for any practical use of these tools, which was less of an issue for the (admittedly simpler/more scoped) topics in earlier courses.