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

4.9

星

52,522 個評分

•

5,940 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....

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.

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.

篩選依據：

創建者 Amit K

•Dec 04, 2018

This is good course for the student, who want to do real stuff with NN. Some of the tricks are well explained like L2,dropout, adam, momentum, minibatches etc. I think these are much needed tricks if i need to implement and tune my own NN on my own problems. I prefer to have a second level of such course which really talks about challenges in real life NN and how to solve those. Once again thanks alot for the entire Team for pulling this together.

創建者 Eleanna S

•Mar 04, 2018

Very useful course. Gives great insight on the hyper parameter tuning, regularisation and optimisation. One request I have is to provide a docker image which we can use to run the exercises locally. Sometimes I found it hard to build the environment where I can run the coursework. Some of the installations are clashing and it is not clear what versions of libraries are used in the coursework environment. It sometimes requires unnecessary effort.

創建者 Hugo v d B

•Sep 26, 2017

In the second course of the Deep Learning specialization Andrew gets deeper into the different subjects of Neural Networks. Again he does a great job in explaining both the math and the way you can improve the outcoming of deep neural networks. The quizzes and assignments where helpful and not difficult at all. He also shows some good frameworks to work with and gives a nice introduction to Tensorflow. I'm looking forward to start with course 3.

創建者 Parab N S

•Aug 25, 2019

Excellent course demonstrating the ways to improve the accuracy of the deep neural networks. It had been the case with me that I could create an initial model easily, but getting an expected level of accuracy was difficult. This course has made it much easier for me to improve th performance of my deep learning models within a short span of time. I would like to thank Professor Andrew N.G. and his team for developing such a wonderful course.

創建者 Xizewen H

•Oct 06, 2017

This course is where the specialization really distinguish itself from Udacity's deep learning nano degree program -- the model fine-tuning part is very important and there are lots of details can be talked about, but Udacity somehow avoided going into details for it. After taking the Udacity's course first, I feel this course really helped me refreshed some knowledge I learnt as well as teach me much more. Definitely recommend this course!

創建者 Ivanovitch S

•Feb 29, 2020

This course is a bit more hard than the first one. I recommend using paper & pencil in order to reproduce all the equations. I gave five stars because the all material is very well described, however, the last part of week 3 must be improved, mainly that related to the practice assignment. There is no link between the Batch Norm and hyperparameter tuning with to practice assignment. Additionally, TensorFlow 2.0 should be introduced too.

創建者 Ayush K J

•Feb 10, 2018

I will recommend this course to beginners in deep learning. As this course has helped me learn about following topics.

Bias/Variance tradeoff, Different types of regularization methods, Code optimization techniques to speed up learning weights, Different types of weight optimization algorithms , About Hyper parameter tuning, Method for normalizing activation as batch norm, About Multi class classification and An introduction to Tensorflow

創建者 Marcio R

•Nov 15, 2018

Excellent course overall. The explanations given are very intuitive even for complex concepts. The teacher always made sure to ease out any concern that might appear if the topic being discussed is not fully grasped yet. I believe that this is a very important step given that MOOC courses should be open for every one, every person has a different learning rate. I highly recommend this for anyone looking to delve deeper into NN and DL.

創建者 Arun S

•Apr 13, 2019

This course helped me to understand the practical aspect of NN. Tuning of Hyper parameters, Regularization , Algos like ADAM are important for fast and accurate training. I hope i could make use of information in future. However this course gives very little introduction to tensorflow and somewhat doesn't satisfies students i believe. Prof. Andrew Ng gives a fantastic lectures covering all important aspects in details with patience.

創建者 Teyim, M P

•Jan 01, 2018

Though the course was mostly theoretical in content, I believe it taught some of the most important concepts in any machine learning undertaking - making the system achieve higher accuracies. Although I found the course content too compact and things kinda move really fast, I think going through the videos a second time even at a 2x speed would make it all stick. In all, it was a tremendous course. I love Andrew Ng's teaching style.

創建者 Yonas T

•Oct 28, 2017

An excellent class and loved the tensor flow tutorial. One thing I would also like to mention is the fact that Andrew made us do the algorithm coding in the first class from scratch helps a lot to really understand the basics of the neural networks. When you then move to using tensor flow it gets even better. Thanks for whole team, Andrew and all the students around the world who makes the environment/forum so vibrant and helpful.

創建者 Jean T

•Feb 07, 2018

Extremely clear and informative about deep learning algorithms per se. The only issue I had is the Tensorflow exercises: since I had never seen TensorFlow before, I lost time guessing the syntax. A more progressive exercise sheet would help get familiar. The point is that, by having to focus so much on the syntax, one focuses less on the structure of the language, so one learns less well the ideas behind the TensorFlow design.

創建者 David R R

•Nov 15, 2017

This course gives you a better understanding of how to increase the performance of your neural network.

There are some video-lectures that are a little harder to understand and maybe boring but, in general, I recomend this course.

Este curso te da un mejor entendimiento de como aumentar el rendimiento de tu red neuronal.

Hay algunos videos que son dificiles de seguir y quizas aburridos pero en general recomiendo hacer este curso.

創建者 Sikang B

•Apr 01, 2018

Clear and practical, this course sets a good bridge from the old NP based programming model to the modern programming models of using Tensorflow and Keras. The optimization methodologies lead to the very useful aspect of ML: hyper-parameters tuning. Though a lot of these hyper-parameters still feel magical, it is super helpful to know more about them.

Suggest to clearly mark this course as a requirement for course 4 and 5.

創建者 Chinmay K K C

•May 10, 2020

I finally got to delve deeper into the intuitions behind the choice of hyperparameters and optimisation algorithms. It is incredible to see how even the smallest of choices can affect our model's performance and understanding the effects of certain choices of hyperparameters on the overall performance of our model will help us make better decisions in regard to how we set up our models. This course was totally worth it!

創建者 Durgaprasad

•Jan 20, 2020

This course builds upon the fundamentals learnt in the first course. By doing this course I have learnt the importance of regularization, and initialization of weights while training a neural network. The course also gives information on implementing neural networks on large datasets and how to methodically choose the hyperparameters. The course exercises are informative and helped me in solidifying the theory learnt.

創建者 David S B

•Jul 24, 2019

Excellent course - my only complaint is that the grader is really finicky about completing the notebooks in a very specific way. Your submissions get rejected in a very cryptic way if you use certain valid TensorFlow constructions, namely you cannot use "Z = W @ X + b", instead you must type "Z = tf.add(tf.matmul(W, X), b)", which I find much more difficult to read. Nonetheless, I think this was an excellent course.

創建者 Heinz D

•Nov 06, 2019

Great course, great instructor and staff. Good speed and good hands-on exercises. Some flaws in the downloadable material and a couple of everlasting corrigenda, but nothing too serious. Integrity control could be enhanced in the TensowFlow assignment. I wish there were not only quizzes at the ends of the weeks but also inbetween or even within the lectures. Looking forward to the next course in this specialization.

創建者 Jaime M M

•May 29, 2019

Very good course as well, although the exercises need some "debugging" there are some typos and errors. I found that the previous courses exercises where too guided, too easy in some points. In this case are more tricky, but not in the correct sense. I would orient a bit more the way of thinking or refer to external sources to get a bit more on track with TF before coding. Nonetheless, all in all, is a great course.

創建者 Renzo B

•Aug 28, 2019

It was a very insightful course. I learned the basic intuition behind the concepts that Andrew Ng explained. For my suggestions, maybe the deeper derivations and meanings behind the concepts could be discussed in video or just a reading material. For example with the maths behind regularization, batch normalization and etc. could be discussed more in depth in a reading material. All in all the course was excellent.

創建者 Mehedi H

•Sep 24, 2017

Very good one. It was great pleasure to learn momentum , RMSProp and then coming to know how to combine them in Adam. Tensorflow example was great. In tensorflow exercise, using regularization can give a boost in the generalization of data which has been mentioned (and I tested it )-but this could have been a part of the exercise.

However, starting to audit the next course of this series. Best of Luck for me !! :D

創建者 Mikhail G

•Mar 07, 2020

Very nice course, worth taking for everyone who is interested in ML/Deep learning, including the very beginners and professionals. I work at the edge of Neuroscience/ML/AI, I have a strong theoretical ML background, but little practice. Even though I was familiar with many of the concepts before taking the course, it was still extremely useful to hear about it again and have way better understanding of the topic

創建者 Ganesh S V M K

•Aug 02, 2020

First of all, I would like to thank Coursera for providing the course. I would always be in debt to Coursera for providing me with financial aid. This website is one of the best online learning platforms. Love the way the assignments are provided. Even I have a bit of understanding and experience in deep learning, this course clears all the blue skies in between and makes deep learning looks simple to learn :)

創建者 Lyle T

•Sep 15, 2017

Very good in-depth coverage of mini-batch, ReLU, Adam, L2 and dropout regularization. Good overview of batch normalization. Brief but useful intro to Tensor Flow (including programming assignment). In general, the programming assignments are pretty easy, but a bit hard to debug in the Jupiter notebooks, though I was able to get things working by inspecting the code to locate typos.

Summary: Highly recommended

創建者 Jonathan M

•May 01, 2020

Builds upon the concepts that were explained in the first course in specialization and Andrew Ng's Machine Learning MOOC and really goes more into depth about regularization and optimization techniques. The introduction to frameworks at the end of the course does a great job of showing how this can apply to other concepts. The programming exercises and course material are great overall and very informative.