IIT Ropar · Computer Science

Deep Learning

Prof. Sudarshan Iyengar · 117 free video lectures

Deep Learning has received a lot of attention over the past few years and has been employed successfully by companies like Google, Microsoft, IBM, Facebook, Twitter etc. to solve a wide range of problems in Computer Vision and Natural Language Processing. In this course we will learn about the…

▶ Play on Open Lecture HallCourse page ↗

Lectures

  1. Biological Neuron
  2. From Spring to Winter of AI
  3. The Deep Revival
  4. From Cats to Convolutional Neural Networks
  5. Faster, higher, stronger
  6. The Curious Case of Sequences
  7. Beating humans at their own games (literally)
  8. The Madness (2013-)
  9. (Need for) Sanity
  10. Motivation from Biological Neurons
  11. McCulloch Pitts Neuron, Thresholding Logic
  12. Perceptrons
  13. Error and Error Surfaces
  14. Perceptron Learning Algorithm
  15. Proof of Convergence of Perceptron Learning Algorithm
  16. Linearly Separable Boolean Functions
  17. Representation Power of a Network of Perceptrons
  18. Sigmoid Neuron
  19. A typical Supervised Machine Learning Setup
  20. Learning Parameters: (Infeasible) guess work
  21. Learning Parameters: Gradient Descent
  22. Representation Power of Multilayer Network of Sigmoid Neurons
  23. Feedforward Neural Networks (a.k.a multilayered network of neurons)
  24. Learning Paramters of Feedforward Neural Networks (Intuition)
  25. Output functions and Loss functions
  26. Backpropagation (Intuition)
  27. Backpropagation: Computing Gradients w.r.t. the Output Units
  28. Backpropagation: Computing Gradients w.r.t. Hidden Units
  29. Backpropagation: Computing Gradients w.r.t. Parameters
  30. Backpropagation: Pseudo code
  31. Derivative of the activation function
  32. Information content, Entropy & cross entropy
  33. Recap: Learning Parameters: Guess Work, Gradient Descent
  34. Contours Maps
  35. Momentum based Gradient Descent
  36. Nesterov Accelerated Gradient Descent
  37. Stochastic And Mini-Batch Gradient Descent
  38. Tips for Adjusting Learning Rate and Momentum
  39. Line Search
  40. Gradient Descent with Adaptive Learning Rate
  41. Bias Correction in Adam
  42. Eigenvalues and Eigenvectors
  43. Linear Algebra : Basic Definitions
  44. Eigenvalue Decompositon
  45. Principal Component Analysis and its Interpretations
  46. PCA : Interpretation 2
  47. PCA : Interpretation 3
  48. PCA : Interpretation 3 (Contd.)
  49. PCA : Practical Example
  50. Singular Value Decomposition
  51. Introduction to Autoncoders
  52. Link between PCA and Autoencoders
  53. Regularization in autoencoders (Motivation)
  54. Denoising Autoencoders
  55. Sparse Autoencoders
  56. Contractive Autoencoders
  57. Bias and Variance
  58. Train error vs Test error
  59. Train error vs Test error (Recap)
  60. True error and Model complexity
  61. L2 regularization
  62. Dataset augmentation
  63. Parameter sharing and tying
  64. Adding Noise to the inputs
  65. Adding Noise to the outputs
  66. Early stopping
  67. Ensemble Methods
  68. Dropout
  69. A quick recap of training deep neural networks
  70. Unsupervised pre-training
  71. Better activation functions
  72. Better initialization strategies
  73. Batch Normalization
  74. One-hot representations of words
  75. Distributed Representations of words
  76. SVD for learning word representations
  77. SVD for learning word representations (Contd.)
  78. Continuous bag of words model
  79. Skip-gram model
  80. Skip-gram model (Contd.)
  81. Contrastive estimation
  82. Hierarchical softmax
  83. GloVe representations
  84. Evaluating word representations
  85. Relation between SVD and Word2Vec
  86. The convolution operation
  87. Relation between input size, output size and filter size
  88. Convolutional Neural Networks
  89. Convolutional Neural Networks (Contd.)
  90. CNNs (success stories on ImageNet)
  91. CNNs (success stories on ImageNet) (Contd.)
  92. Image Classification continued (GoogLeNet and ResNet)
  93. Visualizing patches which maximally activate a neuron
  94. Visualizing filters of a CNN
  95. Occlusion experiments
  96. Finding influence of input pixels using backpropagation
  97. Guided Backpropagation
  98. Optimization over images
  99. Create images from embeddings
  100. Deep Dream
  101. Deep Art
  102. Fooling Deep Convolutional Neural Networks
  103. Sequence Learning Problems
  104. Recurrent Neural Networks
  105. Backpropagation through time
  106. The problem of Exploding and Vanishing Gradients
  107. Some Gory Details
  108. Selective Read, Selective Write, Selective Forget - The Whiteboard Analogy
  109. Long Short Term Memory(LSTM) and Gated Recurrent Units(GRUs)
  110. How LSTMs avoid the problem of vanishing gradients
  111. How LSTMs avoid the problem of vanishing gradients (Contd.)
  112. Introduction to Encoder Decoder Models
  113. Applications of Encoder Decoder models
  114. Attention Mechanism
  115. Attention Mechanism (Contd.)
  116. Attention over images
  117. Hierarchical Attention

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

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