Opiniones y comentarios de aprendices correspondientes a Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning por parte de deeplearning.ai

Very Easy.

Very basic

Too Basic

very good

Awesome!

like it

good

nice

best

Good

good

good

Easy

easy

入门难度

Great introductory course. The two instructors provided a nice introduction to the topics.

3 points of feedback, however. 1: The forums need to be monitored more by Coursera staff; there are many great questions (some basic) in the forums that are unfortunately never answered. 2: The grading app needs to be quality reviewed/reworked. I found myself having to consistently delete the last two unnecessary cells in the submitted notebook [something not very well documented]. Also, the error messages from a non-pass submittal are vague and not very informational. What's causing the syntax error in line xx? The syntax is perfectly fine. Code may be pefectly correct, yet fail the grader algorithm due to these quirks. 3: What is an "adam" optimizer and why am I using it? Even if it's complicated, a note about why it's out of scope and we need to use it here because of X would be very helpful for beginners.

Thank you!

The programming assignment submission system needs work. The course content is decent but very unhappy with submission system. It is very challenging to submit. I spent more time on the first two assignments figuring out how to submit than I did on the assignment themselves. I had the correct work -but the submission system stinks. Also there are issues with differing versions of python/tensorflow; I got hung up by slight changes in tensor flow api -- key values 'accuracy' versus 'acc' were challenging to debug. Ultimately I found it easier to first develop the solution on my local computer -- I would get the code running correctly. Then I would copy/paste that into the colab notebook -- but here is where I ran in to trouble -- differing versions of tensorflow. But it wasn't only me -- in fact some of the class examples had the same exact issue with key values 'acc' versus 'accuracy'.

The material is very good and comprehensive and the instructors are motivating and well-versed experts. However, for an INTRODUCTION to TensorFlow this course lacks complete introduction into TensorFlow. The very first exercise just dives into the code and does not explain with a single word how TensorFlow is structured, how the library is build, where to find important functions, what important imports are and so on and so one. You have to copy and paste the code 1:1 to get it running without understanding anything about the framework. This is a really poor approach for introducing such a powerful framework. I would have expected at least one introductory video about Tensorflow, its structure and components and what are the most important modules to work with, where you can find which function and so on (keras in tf.keras as high-level module, important functions in tf.nn to work with NN)

The explanations are good, but there are no graded programming assignments and this makes the course way too easy. There are only automatically graded quizes (multiple choice) and the questions are too easy. Full disclaimer: I already completed the Deep Learning specialization from deeplearning.ai so I guess that is partly why the course seems too easy from me. Still the lack of graded programming exercises is not acceptable given that this is basically a programming course. It's a shame to give this course such a low rating (3 stars) because the professor is good at explaining and the course in general has great potential, still without graded programming assignments I don't see how you can guarantee that the people with the certificate has at least a basic grasp of the programming skills required.

This is a good beginner's course, but needs a lot of polish. The presenter is very knowledgeable, but his accent is severe, and on difficult words the transcript is entirely wrong, so there's no way of knowing what he's saying. Several of the reading assignments were mis-timed, some of the reading assignments either had dead links, or it was not apparent if there used to be a point to them but there isn't one now. The assignments were buggy -- I spent more time debugging errors in the Jupyter Notebooks that were baked in than on the actual assignments. The assignments themselves were overly easy, but the hassle of debugging made the assignments hard to get through -- the "TA"s didn't answer questions in the forums.

Good course content, but I frequently got lost by the organization of the datasets, files, etc. I learned to set up neural networks but I can't, for example, see how to run them on data on my own computer. Data is just magically present during the course.

Also, it would help to have the Laurence's notebooks available somewhere in the course summary. They are accessible in the unit AFTER he has talked through them in a video, but I always want to see them WHILE he's talking through them (not just the image in the video, the actual notebook), not afterwards.

Some course material is mixed up (e.g. MNIST and Fashion MNIST datsets and examples are 'convoluted';-). Also, the performance of the resulting models is overstated. I don't think it is very impressive to make models that still make mistakes such as qualifying a horse as a human (or worse: an attractive woman as a horse). The idea from the media is, that computers/algorithms beat humans in image recognition easily (e.g. recognizing diseases in medical images), but this is not evidenced by the contents of this course.

The course is fine, but quite basic. I didn't like the fact that there was no way to submit any code homework (as in the other deeplearning.ai specialisation). Just reading some code and experimenting with it is not a good way to learn in my opinion. A suggestion: show a picture depicting the NN that is being built in the code so *this* code implements *this* CNN (or whatever) with depictions of the NN structures as in the other specialisation.

The assignments were a bit outdated and a bit difficult to understand due to the automatic checking ... you need to improve the assignments and tests to make them more user friendly and to evaluate more effectively the topic what the student has learnt(it was becoming difficult to pass not due to the code learnt but due to the automation , memory issues, and the evaluation code. But the Discussion forum helped a lot though!

I consider programming assigments could be better. In comparison to the deep learning specialization assigments, the ones we had in these course are really poor in terms of information and clarity on what we have to do. Specially, it is incredible that there are some unfunctional code on the notebooks such as the javascript part to save the notebook, it only caused error for many students as seen in the forum