IIT Hyderabad · Computer Science

Foundations of Machine Learning

Prof. J. SakethaNath · 61 free video lectures

This is a course on Machine Learning that introduces the subject in a formal way with a focus on basic and foundational concepts. A holistic view of machine learning is taken, including aspects of optimization (SGD) that are ignored in classical narratives. Topics that are relevant to modern models…

▶ Play on Open Lecture HallCourse page ↗

Lectures

  1. What is ML ?
  2. Stochastic framework for Supervised Learning
  3. Risk Minimization and Bayes optimal
  4. Stochastic Optimization and ERM
  5. Linear Regression
  6. Linear Regression - Estimation Error
  7. Supervised Learning >> Denoising/Curve-fitting
  8. Linear modelling
  9. Linear Classification
  10. Logistic Regression
  11. Logistic Regression - ERM
  12. Linear Classification - Estimation Error - Part 1
  13. Linear Classification - Estimation Error - Part 2
  14. Gradient Descent - Primal & Dual view
  15. Smooth Functions and examples
  16. Gradient Descent - Convergence to stationary point
  17. Rapid Convergence to global optima in Linear Regression
  18. KL condition and convergence
  19. Fast Convergence to global optima in Logistic Regression
  20. Stochastic Gradient Descent (SGD)
  21. SGD vs GD for Linear Regression
  22. SGD vs GD for Logistic Regression
  23. Model Selection
  24. What is Regularization ?
  25. Linear algebra motivation for min-norm regularization
  26. Regularization for controlling effective dimension
  27. Bias-variance trade-off for estimation error
  28. Equivalent forms of regularization
  29. Regularization in linear classification
  30. Implicit regularization with GD/SGD
  31. Linear modelling in infinite-dimensional spaces
  32. Introduction to kernels
  33. Linear kernels
  34. Polynomial Kernels
  35. Few properties of kernels
  36. Characterization of kernels
  37. Universal kernels
  38. Universal kernel machines
  39. Bayes consistency with universal kernel m/c
  40. Rate of generalization with kernel m/c
  41. Overparametrization
  42. Modern perspective of ML: Minimal vs optimal regularization
  43. The case for latent variable modelling
  44. Modularity in modelling: Deep learning
  45. Modelling with Deep learning
  46. Estimation with Deep learning
  47. Backpropagation-1
  48. Backpropagation-2
  49. Optimization error in 1-hidden layer nets
  50. Implicit bias in 1-hidden layer nets
  51. Optimization error in deep learning
  52. Likelihood estimation
  53. Energy based models, MLE
  54. MLE examples
  55. Generative regression
  56. Generative classification
  57. Discriminative Models and MCLE
  58. Langevin Sampling
  59. Ensembling: Bias-variance trade-off
  60. Ensembling analysis: Bagging
  61. Concluding Remarks

From NPTEL (IITs and IISc) · CC BY-NC-SA. More from IIT Hyderabad.

Open Lecture Hall is free, with no ads and no sign-up. It is not run by or connected to any university; the lectures are the universities' own free recordings, played from where they host them. About & rights