Welcome to this course on Probabilistic Deep Learning with TensorFlow! This course builds on the foundational concepts and skills for TensorFlow taught in the first two courses in this specialisation, and focuses on the probabilistic approach to deep learning. This is an increasingly important area of deep learning that aims to quantify the noise and uncertainty that is often present in real world datasets. This is a crucial aspect when using deep learning models in applications such as autonomous vehicles or medical diagnoses; we need the model to know what it doesn't know. You will learn how to develop probabilistic models with TensorFlow, making particular use of the TensorFlow Probability library, which is designed to make it easy to combine probabilistic models with deep learning. As such, this course can also be viewed as an introduction to the TensorFlow Probability library. You will learn how probability distributions can be represented and incorporated into deep learning models in TensorFlow, including Bayesian neural networks, normalising flows and variational autoencoders. You will learn how to develop models for uncertainty quantification, as well as generative models that can create new samples similar to those in the dataset, such as images of celebrity faces. You will put concepts that you learn about into practice straight away in practical, hands-on coding tutorials, which you will be guided through by a graduate teaching assistant. In addition there is a series of automatically graded programming assignments for you to consolidate your skills. At the end of the course, you will bring many of the concepts together in a Capstone Project, where you will develop a variational autoencoder algorithm to produce a generative model of a synthetic image dataset that you will create yourself. This course follows on from the previous two courses in the specialisation, Getting Started with TensorFlow 2 and Customising Your Models with TensorFlow 2. The additional prerequisite knowledge required in order to be successful in this course is a solid foundation in probability and statistics. In particular, it is assumed that you are familiar with standard probability distributions, probability density functions, and concepts such as maximum likelihood estimation, change of variables formula for random variables, and the evidence lower bound (ELBO) used in variational inference.
Very good. Liked this course a lot, even though I recognize I should have had a better a background before taking it.
A very hard course but I leared a lot from it. Thanks Coursera and the great teachers
The course is very hard but too important and interesting.
A really valuable learning experience. With these courses, I now feel confident that I can apply the skills from the Deep Learning Specialization in a practical setting.
This has been a great course! The lecture videos are clear, concise, and to the point. The assignments are perfectly structured and the feedbacks from assignments are super helpful.
Great intro to the matter I really appreciated it .
Really interesting and well thogut. I wish there were more advanced courses like that
Great course, sometimes assignments are not so easy and slighty different from teached topics.
Very easy understanding, great for getting practice on TF probability
Really good course touching some really recent research in deep learning.
Amazing experience learning this material, encouraged deeper in dependent dives into material, and it was appropriately structured to set you up for success using that knowledge in other fields.
Showing: 20 of 40
This was a wonderful introduction, not only to TensorFlow probability but also to the basic fundamentals of probabilistic deep learning. This is also, to the best of my knowledge, the only MOOC that gives a somewhat holistic overview of this domain. I'd totally recommend this course to any beginner looking to quickly pick up tf probability.
I really wanted to like this class. I got a lot out of the first two, but very little from this one. As others have hinted, this class is really not like the first two, it is much less general.. it does not just require stats knowledge, but is framed for stats people by stats people, using a library written by someone who does not seem to be clear on python or OOP libraries. There is very little discussion about what you are doing or why, instead treating it like an extension of a stats course where you are learning the function names for various combinations of techniques, but is not really an introduction to the mechanics of the library or of the techniques, with lots of links to 'go read this research paper'. Great for academics studying stats, but the utility to a more general deep learning audience (like the other 2 classes) is limited.
it would probably do well as a stand alone class, but feels out of place in this specialization. The actual topics are really fascinating and worth learning, but the framing is a real barrier to that.
but others got a lot out of it, so take my review with a grain of salt.
Taught concepts were highly interesting and the video lectures were really great. They were very clear and pedagogical. Most TAs had pretty good walkthroughs too.
Why I cannot rate a perfect 5/5 is because many of the lab assignments contained warts and/or misleading instructions that need polish. However, as the course organizer has been frequenting the discussion forum and been very quick to adjust instructions when pointed out by participants, I'm sure the course has already improved greatly. My experience was a bit confused at times however, and I wasn't sure if I had done mistakes or if the automatic judge has bugs, at times.
Overall, I highly recommend this course to anyone interested in learning a modern API for generative modelling. I'm sure I'll come back to the more difficult concepts like the autoregressive trick in normalizing flows, and the change of variable formula, and hopefully having done this course will make those concepts a bit more familiar next time around for me.
Thanks for providing this course material! The future is truly now for online education. :D
Only reason I gave 3 stars is it is definitely NOT a 5 week course. It took me a good 2 months (with a week or so of being pulled off to work commitments). It covers a lot of concepts which need a ton of background. It is dependent on the previous two TF2.0 specialization courses but this specialization is sorely missing one more 5 week course in between the 2nd and 3rd course and that is the Statistics Concepts. Without this I felt that I could not take advantage of the great introduction to all the tools in the Prob deep learning with TF2.0 course.
Very good. Liked this course a lot, even though I recognize I should have had a better a background before taking it.
Never recieved my grade. Contacted coursera support. They demanded another motnh of payment for the course or else I will not get my grade. Terrible practice and this is used to punish people that finish their course early since coursera will simply wait until you pay another month before the start grading your paper!
Tensorflow is outdated and no staff support
Great course (really I absolutely enjoyed it), but this course is finished by peer rated project, where you have to wait for an extended period of time and hope, that someone will review your capstone project, otherwise you are stuck. Since this is an older course, it is completely random when it happens and if you want to receive your ceritificate you just have to wait and pay up. WTF....
While the content is good in terms of the lecture videos and most parts of the coding tutorials, there are a number of bugs in the coding tutorials and assignments. Unfortunately, from what I have seen in the discussion forums, there is no support from the course providers to either fix it or reply to feedback. This is unlike the Deeplearning.AI courses which have much better support
Overall the course was great and should get five stars. But week 3 was so poor and frustrating that I cannot recommend this course before this is fixed (the tutor just reading his code very quickly, and the assignment instructions totally unclear - a lot of people seem to simply give up).
Good Course on Tensor Probability. However, I see a couple of opportunities to improve. During the course, mentors are not available, only students are exchanging. Some of the Models explanation is given as Reading material, instead of Teaching.
Unfortunately, the final course is worse than previous two. The problem is - for me and probably most of learners probabilistic deep learning is something new, and many of us don't have strong background in Bayesian statistics. It's impossible to learn anything from coding labs and assignments where they show you some equations and leave you on your own.
For leaner that already have undergraduate-level knowledge of Math, I highly recommend u attend this course. u see. That's the point, this course not only teaches about some elegant and basic formulas but also focuses on how to use these TFP tools in details, this degree of freedom might be extremely useful in future research as well as career.
However, this balance of duality is not perfect for a leaner who had already familiar with it, also not friendly for those who are not.
But still, it approaches to the optimal solution, I would round up it to five. This IC course is great, but you have to consider about the necessity. Hence if wanna enroll this course, just give me a five, ensure you are truly need take that course.
Interesting course, full of information to implement interesting basic statistical models. I would suggest adding the mathematical background to better explain the interfaces. The course requires previous knowledge in VAE, ELBO derivation, and strong statistic theory. Recommended!
A very hard course but I leared a lot from it. Thanks Coursera and the great teachers
Really good course touching some really recent research in deep learning.
The course is very interesting but you need to be willing to invest time on learning or refreshing key statistics concepts (which is fun!
Most of the teachers are good. But there is one particular week where the PhD student goes really fast in his explanations and the programming assignment has some issues. A pity the instructors have not tried to fix the issues (it seems they do not maintain the courses anymore) as it diminishes the learning experience.
Excellent materials, videos, assignments, and the capstone project but all the great things are ruined by absence of instructors. There is literally no support from the course staff and Coursera Help Center even for system-related technical problems.
The course content is really good and the overall idea of mixing theory, coding tutorials, readings and links to articles is genuinely good. Regretfully, mentors will not answer any of your questions. You are on your own. It is a shame.
As a beginner I was a bit afraid of the course content. I've never build a neural network and though I read 4 or 5 books on the topic my Python and machine learning background was a bit weak. Buy the recommended Manning book on probabilistic deep learning. I read it cover to cover and it helped a lot. Also recommend getting a text on mathematics for machine learning to get a grasp of some of the mathematics presented. I learned so much in the specialization 3 courses and highly recommend this course/specialization. Lectures are good and the programming achievable even with average python skills.