Machine Learning Blink 8.3 (optimizing support vector machines using Lagrangian optimization)

Описание: #SVM #supportVectorMachines #classification
You can download the PDF of the lecture notes at: https://drive.google.com/file/d/1yGD_...


When your intuitive guessing of the optimal solution is evidenced by maths.

To learn more about solving constrained optimization using Lagrangian multiples, you can check the following videos by Khan Academy:
Video 1:    • Lagrange multipliers, using tangency to so...  
Video 2:    • Finishing the intro lagrange multiplier ex...  
Video 3:    • Meaning of Lagrange multiplier  
Video 4:    • Proof for the meaning of Lagrange multipli...  

*** Extra resources on Lagrangian optimization:

1. SVM, duality, and Lagrangian:
http://kseow.com/svm/

2. Why do we try to maximize Lagrangian in SVMs?
https://cs.stackexchange.com/question...
3. Why is solving in the dual easier than solving in the primal? What advantages do we get from solving in the dual?
https://www.quora.com/Why-is-solving-...

4. What is the mysterious dual problem?
https://masszhou.github.io/2016/09/10...

5. Why we need to maximize the Lagrangian at feasible optimal parameters?
https://www.quora.com/How-do-you-know...

6. What is the intuitive explanation for the duality in optimization?
https://www.quora.com/What-is-the-int...

7. Lagrangian for dummies:
https://www-cs.stanford.edu/people/da...

8. Duality and geometry in SVM classifiers:
http://www.robots.ox.ac.uk/~cvrg/benn...
https://cs.nyu.edu/~yann/2010f-G22-25...

For more, check:
https://towardsdatascience.com/unders...