Comentários e feedback de alunos de Guided Tour of Machine Learning in Finance da instituição New York University

was good but expect alitle explanation on the finance stuff

Exercises need better explanations and code

The projects are not so understood.

The grading system isn't good.

Assignments were whack...

Great Course!

Prof. Halperin certainly does have a great deal of practical experience on this subject, having worked in the financial sector for several decades. As such, the lecture videos are succinct and informative. (I will issue the caveat that in order to make the most of this course, one should be already be comfortable with linear algebra, statistics, some calculus and the python libraries mentioned in the prerequisites.) The readings are mostly relevant, at times tangential and in some cases completely off-topic (albeit still somewhat interesting). What perhaps makes this course frustrating, as some have already noted, is that some of the code is outdated. The course, having been taught several years ago, uses version 1 of Tensorflow, and so backtesting assignment code on your own Jupyter notebook can be tricky if you have the latest version installed. It's for this reason that I give this course only 3/5. I got a lot out of it, but it would have helped if it used the latest version of Tensorflow.

The assignments and project are very briefly explained. It took me a lot of unnecessary time to figure out what I was supposed to do. Also the discussion forum is inactive and I have a feeling many leave after seeing not anyone respond to their questions. I think there should be one or two dedicated support answering questions at least within 3 days.

The level of the course in general is pretty high, definitely not beginners level, which is fine I guess, but I do find the lectures are at times going very quick and at times overcomplicate. I would prefer an example to start simple and from there to build for a more complex situation. (For example start the bank failure with say 3 main features and show how you can decide to add another one by showing its impact through deviance and multicollinearity and show how you can then decide to add this new feature or not.)

Actually I was finding that course amazing at first, but I gradually became very upset. The notebooks are way too high level and not self-explanatory. The teacher seems amazing by his knowledge, but one are left with the notebooks without knowing what to do, and the lectures only partially help to solve the problems. A lot of search online needs to be done and I don't think that is the spirit of Coursera courses. I was planning to pay for the whole specialization but unfortunately I will have to give up on this course that was very motivating at first...

This course is a compilation of snippets from different courses. There is a large gap between lectures and readings and the labs. There are several bugs in the labs. It took far more time than advertised to fill in the gaps and complete the course. The content is excellent. But some work to better organize materials, more closely align the labs and eliminate some of the coding mistakes in the labs would help make this a 5 star course.

The lectures, as well as the quizzes, are great and coherent. However, the practical assignments, which are supposed to be the moment of cross-checking your level of comprehension of the learned topics are rather frustrating. I believe it would be of great help to future course takers to clearly state your expectations (e.g. through more detailed exercise descriptions) and introducing vital concepts before requiring their use.

Spent more time than expected. And when I tried to access the last assignment, it showed "404 : Not Found You are requesting a page that does not exist!"I understand the professor and other TA put a lot of effort on these courses, but I would say the assignments are not well organized, and more instructions are needed. Really hope the instructors could update/improve the courses/assignments. Thanks.

Good because it gives a high level good overview of ML in Finance, SVM and Tensorflow.

However, Some examples are very easy and some have been made difficult by providing no references. Tobit regression was very vague. No links to proper reference. Neural Network was the example from Geron's Handbook but there were errors in the custom function that was defined.

More mathematical depth is required.

extremely hard to follow, but better than when it originally came out. I had signed up after numerous ML courses and tried to skip to the later courses in this specialization. I got stuck trying to implement some crazy equations. I'm ok with looking up api methods, but the need to look out for reshaping is troublesome because it's inconsistent throughout the course. Overall, hard to follow.

Good overview of ML and some basic applications to finance.

The pace is very good for people with some training in statistics and maths.

The assignments, however, are not particularly clear and with some obvious errors. There's room for improvement in the description of the exercises as well as including some tests to verify that you're getting the correct output.

content of the lessons is quite good, I would give it 5 stars if the assignments weren't so buggy, contains mistakes, unclear instructions, no help from staff/moderator/instructor, technical issues that are not resolved, etc. a lot of frustration, it just feels like the course was rushed to production and they let the students debug it

Course material is good and a rating of 4 stars or more would have been a fair one, if it was not for very poorly designed and ill prepared assignments. The teaching staff really need to step up a level or two for the assignments.

The course content is good and that the only reason, I am still sticking with this specialization.

The assignments can improve a lot. The jupyter notebooks have no clarity in instructions and most of the time we have to struggle to find exactly what is expected from our code.

The specialization has a lot of potential, anchored only by the lack of the quality of the assignments.

It's a good course but the homework is poorly designed with unclear instructions. Moreover, it's better to get familiar with Python before start this course. The suggested book "Hands-On Machine Learning with Scikit-Learn & TensorFlow" is a very good resource.

Great contents. Excellent topic.

But poor explanation especially in coding assignment.

The assignment includes every coding stuff you need to learn in this course. But there is no explanation about it. You can learn theory from prof. But...coding...?

Very straightforward lectures followed by complex notebooks at a significantly higher level. Given that the labs are using a deprecated version of TensorFlow, with regret, I won't be pursuing this specialization any further.

It is a very broad overview of the machine learning topics but very little about the applications in finance. It wont give you a foundation in machine learning nor any useful insights about financial applications...

The course content is a good review for machine learning with a preliminary introduction on TensorFlow 1.0. However the exercises are mediocre, without clear instruction. Also TensorFlow 1.0 is out of date

Cons: overall content is good. Pros: when you release something (software or scientific article) you often do rigorous testing. Why not to do it with your Jupyter Notebooks? I do not understand it.

The lectures were very good, but the assignments lacked supporting material. Also, most of the further reading was behind a paywall or the links had been removed.