返回到 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization

4.9

星

52,380 個評分

•

5,919 條評論

This course will teach you the "magic" of getting deep learning to work well. Rather than the deep learning process being a black box, you will understand what drives performance, and be able to more systematically get good results. You will also learn TensorFlow.
After 3 weeks, you will:
- Understand industry best-practices for building deep learning applications.
- Be able to effectively use the common neural network "tricks", including initialization, L2 and dropout regularization, Batch normalization, gradient checking,
- Be able to implement and apply a variety of optimization algorithms, such as mini-batch gradient descent, Momentum, RMSprop and Adam, and check for their convergence.
- Understand new best-practices for the deep learning era of how to set up train/dev/test sets and analyze bias/variance
- Be able to implement a neural network in TensorFlow.
This is the second course of the Deep Learning Specialization....

Oct 31, 2017

Thank you Andrew!! I know start to use Tensorflow, however, this tool is not well for a research goal. Maybe, pytorch could be considered in the future!! And let us know how to use pytorch in Windows.

Dec 24, 2017

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.

篩選依據：

創建者 Brennon B

•Apr 23, 2018

Walking away from this course, I do *not* feel adequately prepared to implement (end-to-end) everything that I've learned. I felt this way after the first course of this series, but even more so now. Yes, I understand the material, but the programming assignments really don't amount to more than "filling in the blanks"--that doesn't really test whether or not I've mastered the material. I understand that this is terribly hard to accomplish through a MOOC, and having taught university-level courses myself, I understand how much effort is involved in doing so in the "real world". In either case, if I'm paying for a course, I expect to have a solid grasp on the material after completing the material, and though you've clearly put effort into assembling the programming exercises, they don't really gauge this on any level. Perhaps it would be worth considering a higher cost of the course in order to justify the level of effort required to put together assessments that genuinely put the student through their paces in order to assure that a "100%" mark genuinely reflects both to you and the learner that they have truly internalized and mastered the material. It seems to me that this would pay off dividends not only for the learner, but also for the you as the entity offering such a certificate.

創建者 oli c

•Dec 09, 2018

Lectures are good. Quizzes and programming exercises too easy.

創建者 Alan S

•Sep 30, 2017

As far as the video lectures is concerned, the videos are excellent; it is the same quality as the other courses from the same instructor. This course contains a lot of relevant and useful material, and is worth studying, and complements the first course (and the free ML course very well).

The labs, however, are not particularly useful. While it's good that the focus of the labs is applying the actual formulas and algorithms taught, and not really on the mechanical aspects of putting the ideas in actual code, the labs have structured basically all of the "glue" and just leave you to basically translate formulas to the language-specific construct. This makes the lab material so mechanical as to almost take away the benefit.

The TensorFlow section was disappointing. It's really difficult to learn much in a 15 minute video lecture, and a lab that basically does everything (and oddly, for some things leaves you looking up the documentation yourself). I didn't get anything out of this lab, other than to get a taste for what it looks like. What makes it even worse is TensorFlow framework uses some different jargon that is not really explained, but the relevant code is almost given to you so it doesn't matter to get the "correct" answer. I finished the lab not feeling like I knew very much about it at all. It would have been far better to either spend more time on this, or basically omit it.

As with the first course, it is somewhat disappointing lecture notes are not provided. This would be handy as a reference to refer back to.

Still, despite these flaws, there's still a lot of good stuff to be learned. This course could have been much better, though.

創建者 Lien C

•Mar 31, 2019

The course provides very good insights of the practical aspect of implementing neural networks in general. Prof. Ng, as always, delivered very clear explanation for even the difficult concepts, and I have thoroughly enjoyed every single lecture video.

Although I do appreciate very much the efforts put in by the instructors for the programming assignments, I can't help but thinking I could have learnt much more if the instruction were *LESS* detailed and comprehensive. I found myself just "filling in the blank" and following step-by-step instruction without the need to think too much. I'm also slightly disappointed with the practical assignment of Tensorflow where everything is pretty much written out for you, leaving you with less capacity to think and learn from mistakes.

All in all, I think the course could have made the programming exercise much more challenging than they are now, and allow students to learn from their mistakes.

創建者 Xiao G

•Oct 31, 2017

Thank you Andrew!! I know start to use Tensorflow, however, this tool is not well for a research goal. Maybe, pytorch could be considered in the future!! And let us know how to use pytorch in Windows.

創建者 Alexandre M

•Oct 09, 2019

I really enjoyed this course. Many details are given here that are crucial to gain experience and tips on things that looks easy at first sight but are important for a faster ML project implementation

創建者 Matthew G

•Apr 18, 2019

Very good course. Andrew really steps it up in part two with lots of valuable information.

創建者 Md. R K S

•Apr 15, 2019

Excellent course. When I learned about implementing ANN using keras in python, I just followed some tutorials but didn't understand the tradeoff among many parameters like the number of layers, nodes per layers, epochs, batch size, etc. This course is helping me a lot to understand them. Great work Mr. Andrew Ng. :)

創建者 Abhishek S

•Apr 19, 2020

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

創建者 Yuhang W

•Nov 25, 2018

programming assignments too easy

創建者 NASIR A

•Jan 14, 2020

After completion of this course I know which values to look at if my ML model is not performing up to the task. It is a detailed but not too complicated course to understand the parameters used by ML.

創建者 Abiodun O

•Apr 06, 2018

Fantastic course! For the first time, I now have a better intuition for optimizing and tuning hyperparameters used for deep neural networks.I got motivated to learn more after completing this course.

創建者 Carlos V

•Dec 24, 2017

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

Thanks.

創建者 Shah Y A

•Oct 28, 2019

TL;DR: lectures are awesome, notebooks are bad.

The lectures by Prof. Ng are amazing, comprehensive and intuitive. The prof starts from first principles of simple neural networks and goes onto show concepts like normalization, bias, variance, overfitting, underfitting, regularization, dropout, L1 and L2 regularization, exponentially weighted averages, stochastic, mini batch and batch gradient descent, momentum, RMSprop, Adam optimization, batch normalization and intro to deep learning frameworks. He not only gives the mathematical foundations and code implementations of each concept, but spends a lot of time explaining the intuition behind it, so that we grasp the concept well. It's amazing how he starts from decade old neural networks in the first video, and within 2-3 hours of lecturing, he brings us into the state-of-the-art deep learning models. Thank you Prof. Ng!

But the notebooks have many flaws. The lectures don't set you up for the programming needed in the notebooks. The descriptions in the notebooks are lacking proper tutorial in many places, leading the students incompetent for the exercises that follow. Example: Week 3 tensorflow tutorail; the sigmoid function exercise; the description above the exercise doesn't really teach you how to effectively use placeholders and variables. I was confused and had to go through the noisy dicsussion forum. Please fix it, and if you'd really like more constructive criticism from me, contact me yasser.aziz94 (at da rate ov) gee mail dut com. (lol)

創建者 Anand R

•Feb 17, 2018

To set the context, I have a PhD in Computer Engineering from the University of Texas at Austin. I am a working professional (13+ years), but just getting into the field of ML and AI. Apologies for flashing this preamble for every course that I review on coursera.

This course is the 2nd in a 5 part series offered by Dr. Andrew Ng on deep learning on coursera. I believe it is useful to take this course in order and makes sense as a part of the series, though technically it is not necessary.

The course covers numerous tuning strategies and optimization strategies to help seed up as well as improve the quality of the machine learning output. It is very well planned and comprehensive (to the extent possible) -- and gives the student a very power toolbox of stratgies to attack a problem.

The instructor videos are very good, usually 10 min long, and Dr. Ng tries hard to provide intution using analogies and real-life examples. The quizzes that accompany the lectures are quite challenging and help ensure that the student has understood the material well. The programming exercises are the best part of the course. They help the student practice the strategies and also provide a jump-start for the student to use the code for their own problems at work or in school.

Overall, this is an excellent course. Thank you Dr Ng and the teaching assistants, Thank you coursera.

創建者 Hop B

•May 27, 2019

I would rate for this course 4.5, but Coursera's system does not have it.

About the first and second week, explanation about terms in Deep Learning are very good from Prof. Andrew, the preparation for exams is quite good for you to revise lectures. I think programming exercices should be more challenge and more suggestive for students, but it was okay for me after having some knowledge from Machine Learning Course. I suggest you to finish Machine Learning Course before taking this.

About the third week, i expect a lot more about TensorFlow that Mr.Andrew can give me, or maybe more intuiation about it. Moreover, Batch Norm 's explanation is quite hard to understand, because we do not have any programming exercise for it, so I hope teachers can prepare a programming exercises among with the programiing exercise for TensorFlow.

創建者 Sriram V

•Oct 09, 2019

Insights into best practices and directions for common problems make it an one-of-a-kind material for learners. Andrew, as always, has been commendable with his tutor team, the exercises are well cleaned up and in good shape. May be, if some optional tough exercises are given, it will add more value.

創建者 Artyom K

•May 09, 2019

The topics of this course, such as the setting of hyperparameters and the use of tensorflow, are critical topics for me, and in this course they are explained both in lectures and in practical tasks.

創建者 Hernan F D

•Dec 06, 2019

I enjoyed it, it is really helpful, id like to have the oportunity to implement all these deeply in a real example.

the only thing i didn't have completely clear is the barch norm, it is so confuse

創建者 Hugo T K

•Jan 16, 2020

Very insightful. it would be nice, however, if the course had more information about Tensorflow 2.0.

創建者 陈嵘

•Dec 05, 2019

体验很棒，喜欢这种有作业有评分的课程

創建者 Tang Y

•Apr 15, 2019

very practical.

創建者 Ignacio H M

•Feb 16, 2020

I enrolled in this course without taking the previous ones (I have already done an MSc in Computer Vision and Machine Learning so I thought I wouldn't need the others), but the material has been easy to follow and understand. It is really interesting as it helps you understand important concepts such as bias and variance, or why does batch normalisation work. Sometimes Deep Learning can be seen more as an art than a science, and this course is helpful for defining a good strategy when carrying out deep learing experiments.

創建者 kiran

•Jul 14, 2020

The course began from very basics to complex functions, hyperparameter tuning is efficient in building better models, Kudos to Sir Andrew NG for explaining all of them in the simplest way possible. I would highly recommend this course to all interested in deep learning. But I believe that assignments can be made more challenging rather than just filling up the codes with syntaxes. Logic building is very important.

創建者 Harsh V

•Jan 22, 2019

Add more programming assignments to clear fundamentals.