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

4.9

星

42,297 個評分

•

4,513 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.

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

篩選依據：

創建者 Mihai L

•Jan 21, 2018

This course is also interesting. The art of tuning hyper parameters and other optimization techniques are very interesting and nicely explained.

The introduction to Tensorflow and assignment is also interesting.Overall the difficulty is not high but the concepts are really powerful and important ,most scaffollding is done

創建者 Vlad M

•Sep 07, 2018

The course part is overall good.

The last assignment can be improved in two key ways:

The comment # Z3 = np.dot(W3,Z2) + b3 should be # Z3 = np.dot(W3,A2) + b3 - figured this out by myself without help from forums. :)

Also, the Adam optimization is not very apparent in the instructions - searched in the forums for issues.

創建者 Brad M

•Aug 22, 2019

In my deep learning classes in academia, hyperparameter tuning was always "hand-waved" away - my questions were always deflected, or put off. This class answered every one of my questions, and made me more confident I'd be able to implement a DL system in industry, and be satisfied with the results. Very good course!

創建者 Toby K

•Nov 01, 2019

I am working through the DL specialisation. Consistently good teaching style and the programming assignments are suitably pitched for getting the learner to pick up methods quickly e.g. Tensorflow syntax for self-application later. Good course and looking forward to the next in the series. Well done Andrew and team.

創建者 Ankur T

•Nov 21, 2018

word is not sufficient signup and experience it. For a deep learning beginner who already have math background can easily understand concept behind it but for implementation you need to refer extra materials on internet and book too. Andrew Ng explain only concept and recipe but for practice you will struggle hard.

創建者 afshin m

•Feb 05, 2018

This course is continuation and a requirement of the first course. Really like the learning style of how first course and the first 2 weeks of the second course taught neural networks by doing all the math and calculations manually and finally introduced Tensorflow with parallels of what was taught in the class.

創建者 arulvenugopal

•Dec 17, 2017

This is another excellent course in this specialization. I enjoyed the programming assignments. The instructions, tips made Tensor flow coding section to be easy . However, few blocks consumed more than few hours, due to placeholders. logic and the TF documentation is overwhelming. I am proceeding to next course.

創建者 Wei L

•Aug 26, 2017

This course is harder than the previous one. It teaches more details of tuning parameters and optimization in deep learning. In the end it also teaches tensorflow which is really helpful. It's like a programming course, nerally all the commands have been already provided, so it's not hard to get the code correct.

創建者 姜云鹏

•Nov 21, 2017

It is really good and teach me the basic understanding of DeepLearning back propagation and gradients optimization like Momentum, RMPS, Adam finally I learn how to use Tensorflow to train my model.

But there are some mistakes in the assignments and also in the grade so that it costs me a lot of time but useless.

創建者 Vinodh R

•Nov 12, 2017

The course content was excellent. The only issue is that there were some glitches with the grading of the second week programming assignment, in that I could obtain the expected output, but with repeated submissions, there would be (different) sections which could not be graded due to unnamed technical issues.

創建者 Renato L

•Jul 03, 2019

Excellent content and very well explained. Thanks for this amazing course.

The course cover the building blocks of a Neural network. Andrew (and his team) did a great job by organizing the content in an evolving way in which you have the chance to build the knowledge from each piece of a (deep) Neural Network.

創建者 Bryan H

•May 28, 2018

Practical programming lessons, and well-paced enjoyable lectures.

Comments:

Move tutorials on TensorFlow to Course 3, which was the most obscure part of the course. TensorFlow isn't as intuitive as other numerical toolboxes, so spending more time on the foundations of TensorFlow might reduce the learning curve.

創建者 Rob v P

•Oct 02, 2017

This second course in the specialization is really great. I have gained a lot of insight in hyperparameter tuning and the reason why they work (or don't ;-). It is much easier now to understand what models are doing and why we need certain techniques. This is again one of the best courses for deep learning.

創建者 Daniel R B

•Jun 06, 2018

I really liked the course. The forum is very helpful navigating programming errors during the assignments.

A thing to improve would be to get the feedback from the forums to the lectures. Specially in corrections that should be made to the programming assignments that don't match the expected result. Thanks

創建者 Steve S

•Dec 11, 2017

Provided a lot of deeper insights passed over in the previous course in the specialization. Between this course and the previous course, you feel like you have a very solid beginner's understanding of deep learning, but one that is also practical enough and comprehensive enough to start coding on your own.

創建者 Marcin G

•Oct 15, 2017

Andrew Ng is a great teacher and will get you excited about improving deep networks. In this course you will get to know how to increase performance of your network. Essential course for deep networks specialists and amateurs. Additionally you will get to know most influential people befind the technology.

創建者 Shashank S S

•Jul 08, 2019

All possible area of Improving Deep Learning models got covered in detail. I liked the lucid and intelligible way of explanation . Since the topics were vast to cover , I would recommend to get the course extended by 1 week with one more programming assignment on using tensor-flow with a capstone project.

創建者 Vincenzo M

•Sep 11, 2017

This course will becoma a foundamental course for people that aim to work in the machine learning / deep learning area because it presents clearly the recent innovations in the deep learning. For production environment people will probably use open source framework, but this course clarify what is behind.

創建者 JOSHY J

•Oct 03, 2019

Excellent course if you are passioned about Deep Learning. Walk you through the most basics on how to tune the model parameters so that you can reach the highest accuracy for the model. The lecture is simple and well ordered. The TensorFlow introduction part is more exciting. Overall a wonderful course.

創建者 Dimitrios L

•Feb 18, 2018

Excellent course! Not only does it address critical deep-NNs training issues providing a clear exaplanation around why these tunings are needed, but also provides some empirical advices (e.g. on level of importance on the hyper-parameters, typical values etc) that can be valuable when training depp NNs.

創建者 Aaron B

•Oct 28, 2018

The only thing I wish for is a 'live chat' when an instructor is available, a IRC/slack/chat room for students to help each other, or faster response time when posting to the forums. Also the forums are a bit clunky (I don't remember all the reasons why), but the search allowed me to find useful posts.

創建者 Shashank M

•Oct 10, 2018

This course offers a very quick introduction to methods that could be used to improve usage of deep nets from a practitioner's perspective. Although the mathematical details are not covered in depth, the material furnishes concise list of topics that could be researched upon for in-depth understanding.

創建者 Sachin G W

•Dec 11, 2018

Amazing course, starts right off the bat with hyperparameters, regularization and tunings.

Studied about various optimization algorithms and normalization alongwith mini batches, also the TensorFlow framework.

Thank you to everyone involved in making this course. I highly appreciate what you've made us.

創建者 Muhammad s k

•Dec 03, 2019

I always held an opinion that highly qualified instructors, specifically those holding doctorate degrees are not the good teachers because they can't teach students at their levels. But Sir Andrew Ng proved me wrong, he is a wonderful teacher and tries to explain the minute details.

Salute to you sir.

創建者 Edoardo S

•Jan 20, 2019

Very impressive course, really well done and interesting. One suggestion: apart from the modelling part in the programming assignment, I would also introduce some coding about the computing of the results and the final cost plot (in all the programming assignment these parts are already pre-compiled)