學生對 Google 云端平台 提供的 特色工程 的評價和反饋

Many installations on the pylab notebooks are broken. There are version conflicts all the time. Even if I run the notebook with no changes of my own, several errors appear. This wastes a lot of time.

some issues on some labs due to OAuth compatibility ...

see

ERROR: witwidget 1.5.1 has requirement oauth2client>=4.1.3, but you'll have oauth2client 3.0.0 which is incompatible.

Please update instructions. UI has been changed.

Some code doesn't execute. Last lab. Should be updated. This can be just one sentence (simply, versions of packages don't fit).

This wasn't a bad course, but it is more geared towards showcasing GCP features (BigQuery, Dataflow, Apache Beam, etc.) rather than teaching feature engineering.

The course focuses much more on the gcp tools rather than the feature engineering, labs were not easy to follow, some pieces of code did not work properly.

A good course overall. However, the last two labs didn't run since packages couldn't be installed. Please update these labs. :)

Some lab notebooks need to be updated. Especially for week 4, some setup steps are missing. Otherwise, good content.

Some of the Labs do not work and the information provided are not enough to debug the issue.

Too long for one week. I would suggest to split it in two or even three weeks

The work needed was waaaaay below a one week

Some missing steps in lab descriptions

Some labs don't work

I feel that this, and the tensor flow course that proceeds it in the specialization, were a waste of my time. My feeling is that this entire specialization is a glorified demonstration of what GCP can do with ML. The labs are not interactive and in some cases did not work. I don't feel that I have learned anything new. If I were to use GCP for ML purposes, I would need additional training to do it. I don't recommend this specialization.

Great instructor but (1) the coding challenges are buggy and don't really teach you anything and (2) a lot of the material in this course is tedious for someone with professional training in AI but no experience with GCP

The content is dense but taught superficially. The answers are given away and students have no time to explore the content. The lectures should be broken down into more weeks so that students can absorb the information.

the lectures are good, can be boring. The course would have been more interesting if it had thought-out assignments instead of demo-code to just run as labs

The gap between the lecture and the coding is too big. The coding sessions need to be more interactive to be useful.

Basically this course would receive four stars, but repeated problems with qwiklabs had a severe impact on my overall experience. I got thrown out three times in a row (and my account locked) during dataflow lab.

Every time I had to request unlockin of my account, which took half a day every time. When requesting advice to avoid this error, I got offered the general and vague explanation that I "should only use the resources required by the lab". I am 100% sure that I didn't use any extra resources, including zones and regions.

The Coursera's helpdesk went behind the excuse that Qwiklabs is a third-party service. That may be the case, but since Qwiklabs has been integrated into the Courseras' course, the ultimate responsibility lies with Coursera.

I hope that Coursera will co-operate with Qwiklabs to sort out this very annoying problem.

Ultimately I found this course to be disappointing, because the Google APIs for DataFlow, BigQuery, etc. are unusable with the provided QuickLabs account. When you try to activate any API during the labs, it asks you for a location. It is a required field that says: "You must select a parent organization or folder." Clicking this option reveals a single organization called "no organization," which is not a legitimate choice. APIs cannot be activated and then cannot be used in the lab.

Because of this I was unable to actually do many of the labs that required the use of the Google APIs including the keystone lab "Improve ML model with Feature Engineering" where the taxi-fare prediction model is refined into a perfected state.

I'm upset that I paid money for this.

The last three sections of this course are very difficult. I think the material needs to simplified, less prepositions, to much explanation not enough demonstration, use a thousand words to explain straight forward concepts makes the last part of this course impossible. If any one completes this section with a clear understanding of it's fundamentals, I wish they'd give me a call - frustration - aargh!

This module is interesting but unfortunately it is also plagued with problems. Two key issues that hopefully can be looked into. Firstly, there could be better explanation on Apache Beam. Secondly, I can't run quite a number Qwiklabs because modules not found or some simple import commands are missing

Labs vms are to slow. Speaker is difficult to understand. Mic varies and speech pattern is not clear. The presentations need some graphics rather than a guy talking. Sketch out the ideas on a white board rather than talking 5 minutes to a single slide.

The course needs to be cleaned up. Quizzes have typos/unclear questions; labs ask for too much or not enough; there are lab intro and solution videos for labs that don't exist. Forums seem to be inactive as well.

this is useless...google is advertising their product and making us pay for it. They should learn dr Andrew Ng and create courses which teach us without using a specific platform.

Pros: Tensorflow is an excellent framework for deep learning

Cons :

1- The way this material is designed is 10 X SHIT

2- Either teach properly or don't teach at all.