yes

fun

OK

bu

..

11

NA

T

Very good course, but I did have some problems with how the instructors recommend reproducible splitting the Train, Validation and test datasets. By splitting on a date-hash you are not choosing a random sample. For example, in both the airplane and taxi example, If more winter dates were samples, we would expect more delays and longer taxi rides (and thus more expensive). Wouldn't it be better to split randomly, but reproducibly? In both R and Python SKlearn library you can 'set the seed' when splitting the data which seems like a much less biased and just as reproducible way to split the dataset. The other constructive criticism is to not give all the answers right away to encourage us students to actually write the SQL code (or at least part of it) ourselves.

This course provides a great synopsis of different machine learning models and their nuances. If you haven't seen machine learning before, you will probably need to go slowly and look up some of the concepts on your own. There are a lot of ML terms thrown around without any explanation, so be ready for that if you're new to this. The best part of the course was the Google TensorFlow Playground, where you can experiment with tuning neural networks to classify different types of datasets. The speakers in this course are all very good and the material is well organized. The reason for giving 4 stars is that the quizzes and lab exercises were much too easy, so anybody can get 100% for this course, which makes the grade (and passing score) meaningless.

An interesting and short but intensive course. It introduces a lot of new (to me) tech such as Tensor Flow and Big Query. I learned a lot in a short time, but felt that if I hadn't already had a bit of a grounding in ML I might have been lost. During the course there were a few references to it being part of a specialisation, but I couldn't find what this was and it was not made clear before I signed up that this was the case. Perhaps that is why in the beginning it felt a bit like coming into something half way through, Overall then, interesting and useful, but would benefit from a bit of a clearer setup and explanation of how it fits into the overall Google cloud catalogue.

This course gave me a practical understanding on how machine learning works , how an ML model can be optimized with minimum error and enhance the performance of the model in a better way . I like to thank Google Cloud Team who has taught this course in very interesting way and I'm looking forward to learn the next course of this specialization.

wish they teach you more of the programming side of things and knowing exactly what and why to upload different libraries and or show us how to build these in the labs. Not the first ML course, I've taken but some new people may be a bit confused on the python / setting up / sql even if they have a general knowledge to python and sql.

Excellent course content. Would be 5 stars if the labs forced you to think about how you could apply your theoretical learnings. Unfortunately, labs already have all the code populated, so you just end up running things with the illusion that you have understood everything. Give us labs that require us to solve a problem!

Very useful intro to data processing, specially the hashing mechanism to partition the datasets.

The last lab was confusing because the data might have some invalid value. in the jupyter notebook, the sin, and arcsin values were not getting computed (probably?) as I got warning from python .

I was initially considered that it was too mathematical. But you really don't need to understand the minute details and just get the concepts. good for someone like me that doesn't intend to code but be able to understand enough of challenges and the process for developing models.

Dear team,

I would say the course was really useful and knowledgeable. I like the laboratory parts more. I think google have to consider and concentrate more on lab sections. Overall the course is aewesome.

Thank you for providing the course.

Regards,

Dhanraj S

Lesson Learnt: Best model might not be a good model in real world! Generalization is important!

The labs had issue of disconnection. My jupylab notebooks were frequently disconnecting from the server and I had to manually reconnect them to kernel.

Great introductory to Machine Learning, although the history part is a bit overwhelming for me because he is using a lot of jargon and there is a lack of visualization. But, the rest of the course is great, especially about the sampling.

Little challenging than the first part of the specialization but thoroughly enjoyed deep diving into understanding basic concepts of Machine Learning without being overwhelmed. Great for a person who does not have any previous knowledge.

Will be good to dumb it down further. The last part is good, the first two parts can have better examples and find easier ways to explain the theoretical concepts for folks who have not heard these before.

Tensorflow Playground is awesome to understand some of the theory of Deep Neural Nets !

Theory on creating/managing models was good too.

The labs with BigQuery were not that interesting too

Interesting course, and the technical details during week 3 and 4 were highly appreicated. The labs could have been a bit better put together, but all round happy with this course!

I got a whole idea on how to work on data from scratch. Model selection, generalization, splitting of data and performance metric were few things I learned from this course.

Great presenter. High energy engaging. The material is more difficult and to develop intuition of why the sampling needs to result in constant RMSE didn't come across.