30 июня 2019 г.
The course is very good and has taught me the all the important concepts required to build a sequence model. The assignments are also very neatly and precisely designed for the real world application.
29 окт. 2018 г.
The lectures covers lots of SOTA deep learning algorithms and the lectures are well-designed and easy to understand. The programming assignment is really good to enhance the understanding of lectures.
автор: Yuxiang L•
22 окт. 2021 г.
I learned what this course says it to be. There are many interesting topics and new ways to do things. This course is quite advanced. If you have been in similar fields you may find a lot of connection, such as, time encoding in the transformer networks. You will learn how to incorporate time axis, time dependency, auto-correlation (you name it) to formulation of a machine learning problem using neural networks. I highly recommend this course if you need a systematic treatment of recent advances in speech recognition, machine translation, natural language processing, and more. The padding and blank words also are interesting way to pre-process or post-process. You will earn a lot of bells and whistles too.
автор: Nishant M K•
7 июля 2021 г.
Great introduction to sequence models! Andrew as usual goes into detail on some seminal architectures that have shaped deep learning sequence models over the past ~decade. One feedback: I wish there was a week dedicated to doing just backprop and gradient computations for plain RNNs as well as LSTM-cell based RNNs. The latter is covered to some degree in the programming assignment as an ungraded part, but I don't think that is enough justice to the topic. Also, another feedback: sometimes keeping track of the dimensions of the various entities in play was very difficult for me. Perhaps 10-15 minutes dedicated to explaining just that would be helpful to students. All in all, a fantastic course though!
автор: Adrian N K•
14 февр. 2019 г.
It was an unbelievable journey through this Deep Learning Specialization! I really felt the power of the tools I obtained during the past 3 weeks that it took me to pass all 5 courses of the specialization. Many of the Programming Assignments are demanding and in the end I could be extremely satisfied that I succeeded in taking them all. Thanks a lot to Andrew Ng and all involved for making this sequence of courses accessible to people like me, and presenting it in such an understandable and interesting way! Now, I can start thinking of the vast potential for using Deep Neural Networks not only in Research and Space Sciences, where my interests are, but also in my daily life. Very many thanks again! AJ
автор: Maurice M•
8 июня 2020 г.
The whole series was excellent but in particular this last course on RNNs. Thank you for not skipping the mathematical details and letting us figure out backpropagations through time and how Adam works under the hood and explaining LSTMs and Attention so well. There was even a notebook on Attention! And the dinosaurus notebook was cool but the jazz improvisation really blew me away: the music actually sounded really nice! :) Also, thank you for pre-training the models to safe us time and teach us how to resume training from learned weights! The quizzes were helpful in developing an intuition and the price point was more than fair. Perfect series, Andrew, thanks a lot!
автор: Stephen M•
9 нояб. 2020 г.
Another excellent course, well presented, with compelling content. My only concerns are regarding the labs. With no previous Python or Keras experience, I found I needed to spend a lot of time coming up to speed on new programming domains in order to complete the assignments (my previous experience is mainly C). While this was somewhat an issue in the previous courses in the Specialization, I found it particularly so in Sequence Models. This distracted from the main objective of understanding the core NN algorithms. I would recommend either: 1) advising students to have a solid background in Python, or 2) a bit more clarity on how to use the Keras functions in the labs.
автор: Francis S•
26 авг. 2019 г.
Previously, I have taken online classes before in Machine Learning by going the cheap route (Udemy, blogs, youtube) and you get what you pay for. Andrew Ng explains it the most thorough, easiest, and simplest way possible. Presentation material is very understandable. Great class for new machine learning learners. Highly recommend it. The only downside is that the programming exercises are little too easy in my opinion. I feel like the best way to get your hands dirty is to do actual projects (do your own projects). These lectures are good for intuition and background of different types of Neural Network architectures. Other than that, Great material. Thanks Andrew!
автор: Hermes R S A•
18 апр. 2018 г.
A very good course. It presented gated units like GRU and LSTM with so much simplicity that anyone can understand it on the first run. The downsides were the Jazz music generation, since it was the only task where the data is non intuitive (MIDI files) so you black-box apply the algorithm to a data you have no idea how it is structured, unless, of course, you are familiar with MIDI files prior to this course. Other than that, the learning curve was a bit slower in the beginning, but explodes by the end of the course, where you put all the subjects you've learned to perform a neural machine translation, which, in my opinion, was hugely awesome and rewarding.
автор: Dipan M•
15 июля 2018 г.
Like all other course in this specialization, this is also indeed a great course. It fundamentally clears concepts and gives very clear concpts for topics such as RNN and LSTM, which can ohterwise can be difficult to digest. Also, the programming excersices, built on great topics, suh as Music synthesis, Trigger word activation, are exciting to work on. The only feedback I would like to suggest, is that topics of Backpropogation for sequence model is critical and should have been taken up indepth in study rather than left to excerciss only. Overall this course is more fast paced and packed 3 weeks which should have been perhaps a 4 week course.
автор: Shuvayan G D•
30 июня 2019 г.
This course teaches in-depth knowledge of sequence models in natural language processing and speech regocnition . The programming excercises and the quizzes provide more content to furthur your grasp on the matter . The progamming exercises being totally in Keras , provides a clear analogy of how LSTM s and GRU s , work along with attention models introduced in the last week. You also have to implement a LSTM and RNN from scratch in Numpy , which provides for the basic knowledge how these architectures actually work. Overall , it was a great experience and taking this course should be a pre-requisite for all learning in NLP.
автор: Jeffrey S•
27 апр. 2018 г.
Whew! This was very interesting and challenging. I have a huge backlog of things I need to go back and read up on and better understand. I really appreciate the work that Andrew and his team put into these courses. The lectures were very well paced and clear. His temperament is exemplary for a teacher and his subject knowledge comes across. I found the exercises really well thought out and beautifully crafted. The coding style could not have been more clear and the consistency made it understandable despite the complexity of the subject and the limited time to delve into the mechanics of Keras and the Python tools. Bravo!
автор: Matthew C•
28 мар. 2018 г.
The last course is in this series does not disappoint. I found this course to be more difficult than the others; likely because I had very little prior exposure to recurrent neural networks. However, this course is worth the effort as it opens up a realm of new possibilities; text, audio & time-series data. Whether you need to detect, classify or translate sequences, or even generate new sequences in the vein of some examples, this course is for you. There are several high-level APIs for performing these tasks but having a deeper understanding of what these APIs are doing is invaluable to your success. Take this course.
автор: Ricardo S•
4 мар. 2018 г.
An extremely well thought off and comprehensive introduction to sequence models, with examples taken from the most important/interesting application domains. Andrew NG's clarity of exposition is absolutely wonderful on such an otherwise complex area. The assignments are very cleverly chosen and helped me to finally get to grips with Keras. This being a new course, the assignment notebooks had a few minor issues that are well known by now and documented in forums and erratas, and will likely be fixed in subsequent reruns. Nevertheless, given the breadth and quality of the content, 5 starts are absolutely well deserved.
автор: Mehran M•
22 июля 2018 г.
This was, in my opinion, the best of the 5 courses. Actually, here's how I'd rank the courses (from best to worse):
5, 1, 2, 3, 4
I learned a lot about sequence models and half-way through the course, I was able to jump right in and try some ideas I had in PyTorch.
The assignments could use a bit more work: I didn't really feel inspired by them and their "fill in the blank" style prevented me from thinking too hard.
All in all, I highly recommend this entire specialization. I was completely clueless about deep learning at the beginning, but now I'm actually trying out some novel ideas!
Thanks so much Andrew and the team.
автор: Rahul K•
19 мар. 2018 г.
This course, undoubtedly, has the toughest assignments compared to all the previous courses. The content is rich and informative. Again, pay close attention to the hints given in the programming exercises. If you don't follow, check the Discussion Forums to get a hint. Professor Andrew, your teaching is absolutely sublime - Crisp and concise. Personally, I would have loved an entire week dedicated to Attention Models as the entire concept seemed a bit rushed. Other than that, I have absolutely no qualms! For the people who are enrolling for THIS course only - make sure you're pretty good with Python and Keras.
автор: David R R•
20 февр. 2018 г.
Such a great course. It explains everything from scratch and teach you how to code in numpy (scratch) and how to code in keras to build high performance system (instead of tiny datasets).
I recommend this corse and the DeepLearning specialization as well. Thank you.
Es un curso muy bueno. En el se explica todo desde cero y te enseña como programar los modelos en Numpy (desde cero) o usando keras para crear modelos de alto rendimiento (a pesar de los datasets pequeños por falta de capacidad de computo).
Recomiendo este curso a todo el mundo asi como tambien las especializacion completa en DeepLearning. Gracias
1 мая 2018 г.
This a the last and the most anticipated course for me. It's hard, informative and most useful. I've got chance to learn some popular and powerful methods within the years, like word embedding and attention mechanism. I start to understand the way deep learning community deal with NLP, i.e., ingenious design of network structure inspired by the pattern human beings perceive the world. It doesn't enjoy solid foundation as statistical learning does, but is works and suitable for engineering. That's astonishing! I hope I can combine deep learning with traditional methods to better understand NLP.
автор: Boyko T•
14 сент. 2020 г.
I just want to say Thanks to Andrew and the team for a great content. I may not be able to create award winning NLP models after this course, but I have learned a bunch about them. Lots of work went into creating great videos and even more in creating the programming projects. I really appreciate the format of the programming assignments. For someone with not much experience in DL, they were pretty close to perfect: I felt I was not left to fend for myself, yet they were not overly simple and forced me to solidify what was thought in the lectures and learn better. Thank You Andrew and team!
автор: Hu H•
3 янв. 2019 г.
Thanks very much for Andrew Ng and the other teachers, who made a series of these awesome classes including videos or programming works running on the jupyter-notebook. And also thanks the finical aid provided by the Coursera, I can't finished this course without your generous help. After a hard work with the Deep Learning classes, not only gained the knowledges, but inspired by the spirt from Andrew that "try to help people with your technology", which actually changed my mind, I will study more, do better to remember that in my life. Thank you and hope the world be a better place.
автор: Adarsh K•
19 янв. 2020 г.
Awesome Course! Learned a lot. Would highly recommend this to anyone willing to learn NLP, Sequence Modelling, Word Embeddings, Machine Translation and related stuff. The course builds from fundamentals of NLP like RNNs then LSTMs/GRUs to Word Representations to Sequence-to-Sequence Modelling. At the end you'd learn so much that by just looking at a single slide of an overview of Trigger Word Detection you could make the entire DL model yourself. You'd be fluent with Keras after completing this course. I'd like to thank the Instructor, the Teaching Assistants and the mentors.
автор: Willard C T•
1 июля 2020 г.
I have taken now 6 or 7 courses conducted by Andrew Ng, including this series of 5, and it is absolutely amazing to me that a person of his eminence & level of achievements would even take the time to offer courses like this series. And, what makes it still more incredible is his sincerity, humility and genuine enthusiasm for the subject matter and his gift for explaining it, especially when it becomes very complex. It is just so inspiring; he is truly a rare & exceptional person & teacher and I look forward to taking whatever other courses he is conducting or recommending.
20 февр. 2019 г.
The last course is a little bit more difficult than the previous! Although I majored in Civil Engineering and got my Master's degree in 2018, since I finished the Machine Learning class of Ng 2 months ago,I found this art is so charming and powerful ,so I continued to finish the CS229, That is also a wonderful course!! And today,this DL course was also completed, now I am attending the CS231N class~ Thank you Ng ,thank u cousera, because of you,I have a chance to attend those amazing course from the most famous university. Ng,thanks,you are doing a great thing,thank u!!
автор: Adnan L•
11 февр. 2020 г.
Amazing course. This course was very informative. The assignments gives students the ability to code in keras and use those NLP models described in lectures in the programming assignments.
I felt there was enough help during the programming assignments from the instructors /mentors on the discussion board.
The only thing I wish about this course is to let the students program the Data science part of the programming assignment. I felt some of the details of the pre-processing of the data was already done. It would have been nice to do that or add as an optional part.
автор: Solomon W•
11 февр. 2018 г.
Very frustrating grader. Really time wasting. What is this team trying to accomplish with such disorganized efforts? I hope to see more improvements in the future. I have just completed week1's assignments and revising my reviews from 1 to 4 because the course content is really good and has softened the disappointments caused by the grader.
After week1, the grader frustrations eased as it was working more and more consistently. Most importantly, I learned lots of cool stuff and so I am revising my reviews from 4 to 5. I hope all grader issues are now resolved.
автор: Srikumar K S•
4 июня 2022 г.
Superb instruction mode that strikes a good balance between the nitty gritties of working with language data and actual model building. The only disappointing bit about the latter parts of this "sequence models" course was that the models (esply the transformer models) are too time consuming to train in the context of a programming assignment and so we used pretrained models to check results. This was a bit of a let down (albeit justifiable) since the satisfaction of seeing the transformer model we coded perform still seemed far away (also understandably).
автор: Florent G•
8 июня 2019 г.
A huge thanks for this journey in the specialisation. The material is of high quality and the pedagogie of high qualiber! My only regret is that the course is not longer :P I would have love a course about GAN for example. Also an advanced followup on this specialisation would be amazing. Wanting to learn more i will probably continue my path with https://eu.udacity.com/course/deep-reinforcement-learning-nanodegree--nd893?referrer=nvidia&utm_source=nvidia&utm_medium=partner&utm_campaign=referrerpage, however i would love to continue with deeplearning.ai !