36:07 in the way that it is shown in the slides, I think if kernel-width is large, we are effectively considering smaller region(instead of larger) since the weights will be more damped/smaller now. Correct me if I am wrong.
ig you just compute avg distance to k-nearest neighbors in the training set itself for the training point. so for 1-nearest error is 0 as point is first closest to itself and its an accurate estimation!
Should we always have K as odd value ? This is because we classify x based on the class to which majority of the nearest neighbors belong. If we have even value of K then there is a possibility that equal number of neighbors belong to multiple classes.
I have a question , for new instance we have only x value , so basically (x,0) type of value and it will always be nearest to the lowest y of that x or some( x2,0) type of neighboutr.
You are lil bit confused and that's totally okay, for a new x (Instance ) of course we don't have y and that's what we are trying to find, by applying knn let's say for(k=3) we found 3 nearest points (these points came after applying euclidean distance ) and now the majority points will tell what should be our y. You can also think like that if 3 nearest neighbours have y=1, 2 , 2 then majority is 2 so the y for new instance (x) will also belong to majority class which is 2.
Swathi numbering means what?? you mean like alphabets the way kids starts in order.. m sry plz dont mind but its like that and videos are not published by me.
NPTEL is serving nation in true sense, we should learn from them.
Machine Learning by Prof. Sudeshna Sarkar
Basics
1. Foundations of Machine Learning (ua-cam.com/video/BRMS3T11Cdw/v-deo.html)
2. Different Types of Learning (ua-cam.com/video/EWmCkVfPnJ8/v-deo.html)
3. Hypothesis Space and Inductive Bias (ua-cam.com/video/dYMCwxgl3vk/v-deo.html)
4. Evaluation and Cross-Validation (ua-cam.com/video/nYCAH8b5AQ0/v-deo.html)
5. Linear Regression (ua-cam.com/video/8PJ24SrQqy8/v-deo.html)
6. Introduction to Decision Trees (ua-cam.com/video/FuJVLsZYkuE/v-deo.html)
7. Learning Decision Trees (ua-cam.com/video/7SSAA1CE8Ng/v-deo.html)
8. Overfitting (ua-cam.com/video/y6SpA2Wuyt8/v-deo.html)
9. Python Exercise on Decision Tree and Linear Regression (ua-cam.com/video/lIBPIhB02_8/v-deo.html)
Recommendations and Similarity
10. k-Nearest Neighbours (ua-cam.com/video/PNglugooJUQ/v-deo.html)
11. Feature Selection (ua-cam.com/video/KTzXVnRlnw4/v-deo.html )
12. Feature Extraction (ua-cam.com/video/FwbXHY8KCUw/v-deo.html)
13. Collaborative Filtering (ua-cam.com/video/RVJV8VGa1ZY/v-deo.html)
14. Python Exercise on kNN and PCA (ua-cam.com/video/40B8D9OWUf0/v-deo.html)
Bayes
16. Baiyesian Learning (ua-cam.com/video/E3l26bTdtxI/v-deo.html)
17. Naive Bayes (ua-cam.com/video/5WCkrDI7VCs/v-deo.html)
18. Bayesian Network (ua-cam.com/video/480a_2jRdK0/v-deo.html)
19. Python Exercise on Naive Bayes (ua-cam.com/video/XkU09vE56Sg/v-deo.html)
Logistics Regession and SVM
20. Logistics Regression (ua-cam.com/video/CE03E80wbRE/v-deo.html)
21. Introduction to Support Vector Machine (ua-cam.com/video/gidJbK1gXmA/v-deo.html)
22. The Dual Formation (ua-cam.com/video/YOsrYl1JRrc/v-deo.html)
23. SVM Maximum Margin with Noise (ua-cam.com/video/WLhvjpoCPiY/v-deo.html)
24. Nonlinear SVM and Kernel Function (ua-cam.com/video/GcCG0PPV6cg/v-deo.html)
25. SVM Solution to the Dual Problem (ua-cam.com/video/Z0CtYBPR5sA/v-deo.html)
26. Python Exercise on SVM (ua-cam.com/video/w781X47Esj8/v-deo.html)
Neural Networks
27. Introduction to Neural Networks (ua-cam.com/video/zGQjh_JQZ7A/v-deo.html)
28. Multilayer Neural Network (ua-cam.com/video/hxpGzAb-pyc/v-deo.html)
29. Neural Network and Backpropagation Algorithm (ua-cam.com/video/T6WLIbOnkvQ/v-deo.html)
30. Deep Neural Network (ua-cam.com/video/pLPr4nJad4A/v-deo.html)
31. Python Exercise on Neural Networks (ua-cam.com/video/kTbY20xlrbA/v-deo.html)
Computational Learning Theory
32. Introduction to Computational Learning Theory (ua-cam.com/video/8hJ9V9-f2J8/v-deo.html)
33. Sample Complexity: Finite Hypothesis Space (ua-cam.com/video/nm4dYYP-SJs/v-deo.html)
34. VC Dimension (ua-cam.com/video/PVhhLKodQ7c/v-deo.html)
35. Introduction to Ensembles (ua-cam.com/video/nelJ3svz0_o/v-deo.html)
36. Bagging and Boosting (ua-cam.com/video/MRD67WgWonA/v-deo.html)
Clustering
37. Introduction to Clustering (ua-cam.com/video/CwjLMV52tzI/v-deo.html)
38. Kmeans Clustering (ua-cam.com/video/qg_M37WGKG8/v-deo.html)
39. Agglomerative Clustering (ua-cam.com/video/NCsHRMkDRE4/v-deo.html)
40. Python Exercise on means Clustering (ua-cam.com/video/qs7vES46Rq8/v-deo.html)
Tutorial I (ua-cam.com/video/uFydF-g-AJs/v-deo.html)
Tutorial II (ua-cam.com/video/M6HdKRu6Mrc/v-deo.html )
Tutorial III (ua-cam.com/video/Ui3h7xoE-AQ/v-deo.html)
Tutorial IV (ua-cam.com/video/3m7UJKxU-T8/v-deo.html)
Tutorial VI (ua-cam.com/video/b3Vm4zpGcJ4/v-deo.html)
Solution to Assignment 1 (ua-cam.com/video/qqlAeim0rKY/v-deo.html)
36:07 in the way that it is shown in the slides, I think if kernel-width is large, we are effectively considering smaller region(instead of larger) since the weights will be more damped/smaller now. Correct me if I am wrong.
Hi Ma'am,
U r such a great teacher, ur teaching method helped me a lot.
Thank you so much for making such great videos.
Great Instructor , poor cameraman
Nice teaching
Thank you Prof.Sudeshna Sarkar & Anirban Santara!
Thank you for this great video. It has helped me immensely !
Thank you, Mam you are teaching very good, but lectures are not in a sequence and missing.
Your teaching methodology is simply amazing ma'am.
Lectures on Machine learning in English / HIndi: ua-cam.com/play/PLGeIxG41Dh351Tapkofz0WktplooH5C6s.html
how do we get training error in this, when we're not even training?
ig you just compute avg distance to k-nearest neighbors in the training set itself for the training point. so for 1-nearest error is 0 as point is first closest to itself and its an accurate estimation!
Machine learning made easy with her
Should we always have K as odd value ? This is because we classify x based on the class to which majority of the nearest neighbors belong. If we have even value of K then there is a possibility that equal number of neighbors belong to multiple classes.
If you have 2 class chose odd k and k should not be the multiple of class.
what does "classes are spherical" means?
it means the distribution of data in this case our training data have a spherical form. it is distributed throughout the x-y plane.
Thank you M'am!
Thank you Madam!
I have a question , for new instance we have only x value , so basically (x,0) type of value and it will always be nearest to the lowest y of that x or some( x2,0) type of neighboutr.
Please let me know if my question is unclear the main point we dont have Y then how distance is calculated
You are lil bit confused and that's totally okay, for a new x (Instance ) of course we don't have y and that's what we are trying to find, by applying knn let's say for(k=3) we found 3 nearest points (these points came after applying euclidean distance ) and now the majority points will tell what should be our y.
You can also think like that if 3 nearest neighbours have y=1, 2 , 2 then majority is 2 so the y for new instance (x) will also belong to majority class which is 2.
Great Session
Great Lecture
the way she pronounces ALGORITHM is funny! :D :D
Yep. Because you don't know how to pronounce it.
Ma'am is speaking the finest English. Go check out her profile and you'll know how good she is
@@705pratik9 okay!
@@OmkaarMuley Yep take care man. Always wear a mask.
Plot of decision boundaries using MATLAB:
ua-cam.com/video/uql5RbM9GHI/v-deo.html
thankyou dear mam
You got even the heading of the video wrong Prof. The algorithm is called k-Nearest neighbors!
Text boot or else PDF can you send me link for downloading ur PDF materials please mam. It is useful for my project work
U can download these lecture transcript from neptel site
try to give the numbering for lectures
okk Swathi
is these videos are published by you??
Swathi numbering means what??
you mean like alphabets the way kids starts in order..
m sry plz dont mind but its like that and videos are not published by me.
it does not mean like that .....i mean try to publish in an appropriate orde.....it is better to understand others in a correct way...thank you
No thanks