学生对 密歇根大学 提供的 Applied Machine Learning in Python 的评价和反馈

very good

perfect

good

Good

GOOD

....

Good

Good

.

This is an OK introduction to Machine Learning. It covers a range of relevant topics. The gap between the lecture content and the assignments is the typical chasm for this U.Michigan "speciality", and frankly you end up basing assignment answers more on internet research rather than lecture content.

I'd sum it up as a substantial missed opportunity. The last assignment is really good in terms of doing a realistic Machine Learning project, but the preceding course content doesn't give you the tools or frameworks to do that project in a logical, industry standard workflow. It gives you an idea of what the tools are, but not how to really apply them all together in an efficient and logical series of steps.

It's as if those who designed the course decided that learners needed a tough-love approach, like a trainer lying down on the grass and showing learners swimming strokes, and then just throwing those learners into a pool and expecting them to keep afloat, and combine what they remember with what they see other more experienced swimmers in the pool doing. It shows a fundamental misundestanding of the Coursera learners usually being very time poor and expecting much more from the instructors.

This course provided a good structure and order to learn introductory machine learning concepts in Python. However, I thought the lectures in particular were needlessly more abstract than the previous data science courses in this specialization.

In my experience, learning a new programming concept comes from practically writing code then observing what happened. The earlier data science courses were great because you could test code with the lecturer as the video progressed and learn from it.

The lecture content here structured to discuss broader machine learning concepts, rather than setup to follow along in the notebook. I found this was okay for introducing the idea of different machine learning concepts, though without the practical application and observation it became difficult to remember these concepts or test what I was hearing. I found most of my learning happened in the assignments or by following more practical online resources. The course could be improved by tying the notebook modules more closely to the video content, making it easier for learners to follow along.

I'm glad there was an introductory course like this offered for machine learning. The content is very accessible and the assignments are simple enough to work through without frustration, but challenging enough to help you understand how to apply machine learning algorithms on your own.

I did purchase the book recommended, Introduction to Machine Learning with Python by Andreas C. Muller and Sarah Guido. The lectures in this course are basically paraphrase the book. Frankly, I think you'd get more value from this course if you read Chapter 2 in its entirety and follow along with the juypter notebooks provide with the book. It's easy to tell when someone is teaching you vs. reading to you— this course's lectures were definitely the latter.

I felt like each standalone topic was explained okay, but I didn't get a good big picture understanding at the end. There wasn't a good wrap up to explain holistically how to choose one classification method over another.

There are also just too many mistakes in the lecturer speaking as well as in the slide. For an online class I would expect that Coursera would redo the video or at least the slides as they interfere with learning.

I felt that the lecturer belabored easy points, like calculating precision and recall by hand but then didn't explain other topics regarding the classification methods well. I did not find the graphic visualizations in his slides helpful in explaining hyperbolic tangent functions.

This course was a complete disappointment. First of all, it should have been split into two courses. The second week especially contains so much material to the point that it's not-pedagogical. Also, I regret to say that the instructor is not fit for this task. It would be better if they used Christopher Brooks from the first two courses as he is more engaging and he seems to have a lot more experience in public talking. Another thing is that there are serious bugs with the assignments. This course needs serious redesign.

All in all, don't spend your precious time and money on this one. There are better courses available on this subject.

The quality of the teaching is a marked improvement over module 1 & 2 in this specialisation. In my opinion it would be a 4/5 star course on that alone however there is 1 minor and 1 major issue. Starting small, the course could do with better summary notes/cheatsheets to help remember details and as prompts when doing assignments; I found it really annoying to have to skim read the lecture video transcript or scan through the videos. The MAJOR issue is the problem sets and the autograder. I really feel the teachers need to re-write this whole section before I could recommend this course.

Most of the content professor taught are intuitive, but the PPT seems helpless. Furthermore, the thinks in the course are shallow depth, conversely the assignment are little bit difficult especially on assignment4. I mean if the goal is to train our to do some real world data you may can shrink the dataset, the large dataset would takes more time to training which would cost more time to debug. Anyway, this is a great course but I think it's better to do slight change on the quiz and assignment.

This is a good introduction to applied machine learning with python. Although it is "applied" it would be worth to cover the basics of the presented algorithms a bit more thoroughly. In paticular, I think that the regularizations parameters and their role in bias and variance are not presented in a very clear way. On a different topic I think that the course deserves to be updated with latest sklearn implementations and correct a lot of bugs in the assignements and lectures.

This course helped me to get started on using Python for machine learning tasks.

Personally I would have preferred a more mathematical approach when discussing the various machine learning techniques, in order to learn more about what's going on "under the hood" in scikit-learn. I know that the course is called "Applied Machine Learning in Python", but to me it felt more like "Extremely Applied Machine Learning in Python" :-) Other than that, I enjoyed this course!

There is a lot of good material in this course, but it is noticeably not taught as well as the previous two courses in this specialization. The lesson plan feels like a class lecture modified just barely enough to work as a MOOC, the autograders are highly finicky, and most of the programming assignments had errors or missing details that required the learner to check the class forums to find out how to fix them. Overall, it was a helpful course, but felt unpolished.

I do think the lectures are very well done and I believe I learned a lot. However the programming assignments part was frustrating, there are a lot of issues with the autograder, loading files etc. I would appreciate if the steps were described in greater detail. Some parts are very easy, just blending together a few pieces of code from the lecture, and others very difficult, built on things not covered in the lectures. The last assignment was the perfect balance.

This course is, in fact, excellent. One can learn a number of algorithms used in a machine learning practically. This course does not focus much on mathematics behind tools we used, the professor taught a lot about the practical one. However, some of the parst in this course are too rush; you have to understand a lot of concepts in Python berfore entering this course, including basic Python syntaxes, and practical libraries such as Numpy and Pandas.

The course tries to do too much in four weeks. Consequently, the teaching material isn't as comprehensive as it ought to be. I've probably spent over 10-15 hours cumulatively on other websites, trying to comprehend the intuition behind the algorithms used. This course isn't great at getting that across. There's a lot in here that we're forced to take for granted. I'm afraid I'll have to think twice before I choose other UMich courses in the future.

The material seemed ok. Really annoying that this course genuinely had incorrect code in the homework assignments. It seems that some documents changed directory and were different in the homework folders, vs the grading tool. resulting in failed grades where tests worked just fine. Easily fixed, but why would I have to? Really hurts the notoriety and reputation of this program to have such simple frustrating errors. (on 3 of 4 assignments!)

It would be better if this course was not with Jupyter notebooks. Professional data science projects will not use notebooks but script files instead. The course should prepare students for professional projects by using script files.

Also the lecturing is very rigid and scripted which makes it less engaging. There is also no material on how any of the algorithms work in detail however there is good material on scikit-learn.

While the course material is very helpful and reasonably pace, I felt like I'm always battling the autograder to pass the assignment. I do think that I spend more time to get my answer accepted by the autograder than actually working on the assignment itself. I think an easy way to fix this is to clearly layout the tips to get pass the autograder, rather than having the students to search through the forum for a solution.