Retour à Mathematics for Machine Learning: PCA

4.0

étoiles

1,791 évaluations

•

418 avis

This intermediate-level course introduces the mathematical foundations to derive Principal Component Analysis (PCA), a fundamental dimensionality reduction technique. We'll cover some basic statistics of data sets, such as mean values and variances, we'll compute distances and angles between vectors using inner products and derive orthogonal projections of data onto lower-dimensional subspaces. Using all these tools, we'll then derive PCA as a method that minimizes the average squared reconstruction error between data points and their reconstruction.
At the end of this course, you'll be familiar with important mathematical concepts and you can implement PCA all by yourself. If you’re struggling, you'll find a set of jupyter notebooks that will allow you to explore properties of the techniques and walk you through what you need to do to get on track. If you are already an expert, this course may refresh some of your knowledge.
The lectures, examples and exercises require:
1. Some ability of abstract thinking
2. Good background in linear algebra (e.g., matrix and vector algebra, linear independence, basis)
3. Basic background in multivariate calculus (e.g., partial derivatives, basic optimization)
4. Basic knowledge in python programming and numpy
Disclaimer: This course is substantially more abstract and requires more programming than the other two courses of the specialization. However, this type of abstract thinking, algebraic manipulation and programming is necessary if you want to understand and develop machine learning algorithms....

Jul 17, 2018

This is one hell of an inspiring course that demystified the difficult concepts and math behind PCA. Excellent instructors in imparting the these knowledge with easy-to-understand illustrations.

Jun 19, 2020

Relatively tougher than previous two courses in the specialization. I'd suggest giving more time and being patient in pursuit of completing this course and understanding the concepts involved.

Filtrer par :

par Niju M N

•Apr 09, 2020

This is the final course in the Specialization, that focuses on Principal component Analysis.This course is a bit hard compared to the other two courses in specialization. This builds on the topics explained in the other two courses.The Instructor tries to squeeze the concepts in the limited time.Not all materials are completely explained in the video, however, students can refer to other materials available in the web/ Refer the course forums and get the concepts and use them to solve the Quizzes. Some times the Assignments and quizzes are frustrating , however they do a good job of reinforcing the ideas taught in the video. Totally this is a good time spent .

par Vassiliy T

•Jul 10, 2018

it is good, challenging course. i've learned a lot, but feel that i came away with quite patchy knowledge. This course is a big step up in complexity and delivery form the previous two courses. perhaps my expectations were not right to start with - one cannot learn this level of complexity so quickly. Admittedly there are many gaps between the lectures and course materials and what is asked in programming assignments. i ended up reading a lot online to fill in the gaps (i've learned a lot of python during the course, which is great!).nevertheless, after this course i feel equipped to continue with machine learning.

par Matteo L

•Apr 20, 2020

I think this course is slightly underrated at the moment. The topic is not an easy one and I thought the teacher did a great job of explaining it as clearly as possible using an appropriate amount of mathematical derivation.

I really thought the last week of the course was great, especially considering that everything we had seen so far in the specialization was used to develop the PCA algorithm. It's quite amazing how topics such as eigenvectors, projections and optimization all come together here.

I think the notebooks were quite challenging compared to the previous two courses with is definitely a plus!

par Nikolay B

•Aug 03, 2019

Instructor gives the very dry but useful essence of the "philosophical" concepts of dot and generalized inner product, etc., - personally, liked that. Unfortunately, the offered problems are so far away from the delivered videos but the web search helps on getting the hints. This course makes you think - I learned a lot just by asking myself "what do they mean under this statement?", what they want in this task? Though I will appreciate if providers elaborate the material further and so instead of googling we spend our time watching - a single point access.

par Luke

•Jun 11, 2020

This course was a very interesting end to the specialization. The first two courses teach you the tools necessary, while this course teaches practical (although highly theoretical) application of those tools. The reason for the non-five star rating is the python exercises take one on quite a journey, which often times goes through some dark and murky tunnels which are tough to escape. Also, there are few worked examples during lectures.

par Antonio C P

•Jan 24, 2020

It was a real tour de force on the mathematics, and I had some hard time following the ideas of the instructor many times. However, the topic was completely covered in a very systematic way, which is excellent in my opinion. My only suggestion is to focus more on what really matters: do we really have to spend such a long time discussing about different metrics for an inner product if in the end we only use the euclidean metric?

par Barnaby D

•Jan 03, 2020

Would give this course 5 stars if it was properly described so that expectation could match reality:

Give yourself plenty of time for this course - it will take quite a bit longer than described.

Make sure you are comfortable with Python and NumPy before you start (particularly the linear algebra functions).

It is very different (much less hand-holding) than the other courses in the specialization.

par Nelson F A

•Apr 25, 2019

This course brings together many of the concepts from the first two courses of the specialization. If you worked through them already, then this course is a must. There are some issues with the programming assignments and the lectures could do with some more practical examples. Be sure to check the discussions forums for help. For me they were essential to passing the course.

par Visveswara K M

•Jun 18, 2020

This was a bit more challenging than the previous two courses. I didn't enjoy it as much as the previous courses, however, I learnt more than the previous two. The discussion forums were helpful and the instructors contributed regularly. The assignments were a bit frustrating at times but still manageable. However, the assignments could have had a bit more of explanations.

par greg m

•May 24, 2020

Very good course, interesting material. However the amount of programming knowledge required is way beyond a beginner like myself and I struggled with that , consuming much time. Those with programming knowledge have a tremendous advantage on this course.

There should be a week or a separate brief course on python/numpy.

A follow up more advanced course would be good too.

par Evgeny ( C

•Jul 25, 2018

It was a harder course where I spent double the time I have initially anticipated.

It is much harder than the two predecessor courses in specialization, and amount of direction when it comes to doing exercises is significantly smaller. More Python knowledge is required.

That said, I feel like I have finally understood the PCA and math behind it, which made it all worth it

par Mark S

•Jul 07, 2018

Loved the course, although I wish there was more ramp up to some of the complex scenarios (or anything simple but new). Very helpful forums/community. Requires a fair amount of external reading/referencing for some of the concepts which seem to be covered only at a high level in the lectures.I would love to see more courses on applied mathematics for machine learning.

par Jerome M

•Jul 26, 2018

The best of the 3 courses. This is a refresh course of course. A solid background in linear algebra is required in order to fully understand everything. I personnaly recommen the MIT course from Gilbert Strang before you try this one. The python exercises are very well designed and I can only be thankful to having shared this knowledge. Thank you Imperial College.

par Timo K

•Apr 10, 2018

Not quite as good as the other two courses of the same specialization. Even though the instructor seems immensely knowledgeable he could work on delivering the material (which is more abstract than before to his credit) in a clearer manner.

The programming assignments are great albeit a bit hard to troubleshoot at times. All in all still a great course.

par Joshua C B A

•Mar 11, 2019

Very good course. I liked every single video and exercise. I feel that the programming assignments were a bit more challenging and sometimes I was not too sure of what I was doing. I am not a professional in handling Python, so I had to surf online finding the commands to be able to build the simplest code possible. Other than that, it was enjoyable.

par Cheng T Y

•Jul 08, 2018

good thing is it's trying to give you a sense of practically how to do it.downside is it's not really bridging to from maths to that practical sense in python (and the online jupyter notebook is terrible).the teaching staff is actually more responsive than the other 2 in the specialization.a bit more sided on python than maths though.

par Phạm N M H

•Jul 12, 2019

This maybe the most frustrating course and most advance compare to 2 other courses, you might confuse about the code in the assignment of this course. So, if you do have basic background about coding with numpy, matrices,etc..., I do recommend this course, if you qualify enough to fix the bugs of what the dev team left.

par Thorben S

•Mar 08, 2019

I would have liked to be introduced to the topic on a higher level first - and then, step by step, an introduction of the math to solve specific problems in the progress. That would be a perfect approach, especially for data scientists who just want to understand the underlying math for such a widely used technique.

par Andrés M

•Jul 04, 2020

I believe the course is proper for people that have no prior knowledge in linear algebra whatsoever. I liked how clear it was to introduce concepts, yet I found that if you knew nothing the course is too hard but super easy for the ones that have some knowledge in algebra and calculus.

par Mike W

•Mar 22, 2020

The quality of this course is comparable to the previous courses in the specialization, but the math and derivations were harder to follow (even accounting for the increased difficulty of this course). The assignments also were very practical and help reinforce the course's content.

par Shariq A

•Oct 20, 2019

Thank you professor for providing such a valuable course.

Just I wanted to say one thing without hurting anyone, the week 4 on PCA is not very clear. The derivation are not very correlated .A humble request isthat to elaborate the derivation which would further enhance the learning

par Aarón M C M

•May 10, 2020

I think this is one of the bests courses that I have taken. I would just recommend to describe more accurately decimal precisions in tests because it has a little challenging to realize that the solutions proposed were not successful enough because of this issue.

par Jonathan F

•Mar 17, 2019

This course is way harder than the first two. The maths itself is more difficult. The Python parts are a lot more challenging because they require a good understanding of the way Numpy handles vectors and matrices. But the end result is good and it is worthwhile!

par JITHIN P J

•Apr 27, 2020

Course content is too hard to understand. You need to go through the content at-least 2 -3 times. But its good. Also assignments are bit tricky and you need to do alot of googling which will make you learn more. Thanks Coursera and ICL for this wonderful course

par Moreno C

•Mar 14, 2020

This was the most rigorous and demanding of the courses of this specialization.

The video lectures were well organized.

The interaction with the Jupyter Notebook was sometimes confusing but perhaps this was due to my limited knowledge of Python.

Thank you.

- L'IA pour tous
- Introduction à TensorFlow
- Réseau de neurones et deep learning
- Algorithmes, Partie 1
- Algorithmes, Partie 2
- Apprentissage automatique
- Apprentissage automatique avec Python
- Apprentissage automatique à l'aide de SAS Viya
- La programmation en R
- Intro à la programmation avec Matlab
- Analyse des données avec Python
- Principes de base d'AWS : Going Cloud Native
- Bases de Google Cloud Platform
- Ingénierie de la fiabilité du site
- Parler un anglais professionnel
- La science du bien-être
- Apprendre à apprendre
- Marchés financiers
- Tests d'hypothèses dans la santé publique
- Bases du leadership au quotidien

- Deep Learning
- Le Python pour tous
- Science des données
- Science des données appliquée avec Python
- Bases de la gestion d'entreprise
- Architecture avec Google Cloud Platform
- Ingénierie des données sur Google Cloud Platform
- Excel à MySQL
- Apprentissage automatique avancé
- Mathématiques pour l'apprentissage automatique
- Voiture autonome
- Révolutions Blockchains pour l'entreprise
- Business Analytics
- Compétences Excel pour l'entreprise
- Marketing numérique
- Analyse statistique avec R pour la santé publique
- Bases de l'immunologie
- Anatomie
- Gestion de l'innovation et du design thinking
- Bases de la psychologie positive