Chevron Left
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization に戻る による Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization の受講者のレビューおよびフィードバック



In the second course of the Deep Learning Specialization, you will open the deep learning black box to understand the processes that drive performance and generate good results systematically. By the end, you will learn the best practices to train and develop test sets and analyze bias/variance for building deep learning applications; be able to use standard neural network techniques such as initialization, L2 and dropout regularization, hyperparameter tuning, batch normalization, and gradient checking; implement and apply a variety of optimization algorithms, such as mini-batch gradient descent, Momentum, RMSprop and Adam, and check for their convergence; and implement a neural network in TensorFlow. The Deep Learning Specialization is our foundational program that will help you understand the capabilities, challenges, and consequences of deep learning and prepare you to participate in the development of leading-edge AI technology. It provides a pathway for you to gain the knowledge and skills to apply machine learning to your work, level up your technical career, and take the definitive step in the world of AI....




Very good course to give you deep insight about how to enhance your algorithm and neural network and improve its accuracy. Also teaches you Tensorflow. Highly recommend especially after the 1st course



Exceptional Course, the Hyper parameters explanations are excellent every tip and advice provided help me so much to build better models, I also really liked the introduction of Tensor Flow\n\nThanks.


Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization: 6951 - 6975 / 6,998 レビュー

by Matthew P


Focused a bit on minutia.

by Adam G


Multiple grading issues.

by Chaitanya M


could be more engaging

by Cory N


Update for TF2.0 :)

by Алексей А


Looks raw yet.

by Ilkhom


awful sound

by Akhilesh


enjoyed :)

by zhesihuang






by Long H N



by KimSangsoo



by Maximilian S


T​his is a nice but very basic introduction to the practice of DL (the last week about tf is nice). However, the assignments are way too shallow! In the assignments the students are "spoon-fed baby-food"... one can solve almost all exercises without thinking and without having understood anything (it is mostly solvable by copy&paste).

For instance, I have learned the most in the final assignment when it did not fully work and I forgot a tf.transpose(..) and I actually had to think about what was happening.

A​nybody applying to our group who presents this course as evidence that they know about the contents will not be taken seriously (and rightfully so!) -- thanks for a very quick way to sort out useless applications (anybody presenting a certificate for this course in public).

The assignments could also be auto-graded by using the format of any programming competion (specifying the input-output relation, providing an input, and giving the student total freedom in how to implement the solution), e.g. like in the famous advent-of-code. Then the course would be harder (but way more valuable!) -- however, coursera won't get enough paying subscribers that way I assume.... oh what a pity.

by Sameer C


Terrible construction of programming exercises. They either end up being extremely trivial or vert obfuscated. Sometimes too much information is given with no incentive to think or too little information is given leading to a deadlock. Week 3 of this course is utterly trash. Course content feels rushed and the programming exercise does not explain anything or clear any doubts. Why on earth do I have to do so little in these programming exercises. Why can't you make us write the little helper functions and plotters and the compiled model.

by Sébastien C


Course covers the most important parts of hyperparameter tuning, regularization and optimization.

As a general remark for this specialization, the exercices do not provide any value. We just have to fill in some lines and submit our work.

As I tend to "learn by doing" I had to look for other tutorials and projects on other platforms (Kaggle, MachineLearningMastery's website) in order to complete my learning.

by Fabrizio N


Good course content and clear exposition by Andrew. The course material however is not of a good standard. The slides can be downloaded but after all the hand scribbles by the tutor, they are barely decifrable. Some are just blank pages that need to be filled in with screenshots from of the videos. The assignements are often just a copy and paste exercise, and Jupyter crashes cause frequent loss of work.

by Goda R


The video content is very good to get a good hang of theoretical aspects but the programming assignments are too spoon-fed because of which after doing filling the blanks, you don't feel confident enough to implement the same on your own. Instead the assignments should be changed to cases where instructions are given in words and entire function should be implemented by students.

by André Ø


The TensorFlow part of the course felt out of place and not of the same quality as the previous material. It would have been better if another week was spent using TensorFlow to actually improving a NN and not just copy-paste an example into the assignment. Even after using TensorFlow in the assigment and passing, working with TensorFlow still

by Sergey K


In general the course isn't deep enough. There are no summaries of the lectures, there are no excercises during lectures, the programming assignments are very weak, they don't challenge the use of lectures or anything - all necessary data are in the notebooks. All this course will be lost in a week.

by James H


The whole tensor flow introduction is weak - it clearly requires further reading, which is fine but totally out of kilter with the videos so far, which have taken things from first principles very clearly.

by Martin B


A technical problem with the grader caused my grade to be artificially lower on the last project. Although I was instructed to resubmit, the course ended with a lower grade than I should have received.

by Anne R


Not much implementation required of the students. More testing of the methods would be useful or if the concepts are the focus then this course should be merged into another course in the sequence.

by Brian R


The course material is good but Jupyter notebook interface does not work correctly. You will waste a lot of you precious time fiddling and redoing work that you lose when the notebook fails to save.

by Abhishek K


could be accomplished in a week. wastes time and doesnt go in sufficient depth.

After completion, you will get a 'taste' of optimization techniques, but it is not way comprehensive.

by Don F


The course was good but there were multiple mistakes in the final programming assignment. These mistakes were reported in the forums over 4 months ago and have not been addressed.

by Sam G


I find that the programming assignments have a LOT of copy pasta. Also, wasn't enthused to hear that we are using an out of data framework.