1024 x 768
Deep Learning More Techniques (Lee Hung-yi, 2:29:01)
- 1. Tips for Training Deep Neural Network
- 2. Announcement
- 3. Outline
- 4. Outline
- 5. ReLU
- 6. ReLU
- 7. Review:Backpropagation
- 8. Review:Backpropagation
- 9. Problem of Sigmoid
- 10. Vanishing Gradient Problem
- 11. Vanishing Gradient Problem
- 12. ReLU
- 13. ReLU
- 14. ReLU
- 15. ReLU
- 16. ReLU
- 17. ReLU
- 18. ReLU
- 19. ReLU
- 20. ReLU - variant
- 21. Maxout
- 22. Maxout – ReLU is special case
- 23. Maxout – ReLU is special case
- 24. Maxout - Training
- 25. Maxout - Training
- 26. Outline
- 27. Cost Function
- 28. Output Layer
- 29. Softmax
- 30. Softmax
- 31. Softmax
- 32. Softmax
- 33. Softmax
- 34. Softmax
- 35. Normalizing Input
- 36. Outline
- 37. Vanilla Gradient Descent
- 38. Vanilla Gradient Descent
- 39. Outline
- 40. Outline
- 41. Learning Rates
- 42. Learning Rates
- 43. Adagrad
- 44. Adagrad
- 45. Adagrad
- 46. Contradiction?
- 47. Intuitive Reason
- 48. Larger gradient, larger steps?
- 49. Second Derivative
- 50. More than one parameters
- 51. What to do with Adagrad?
- 52. Outline
- 53. Easy to stuck
- 54. In physical world ……
- 55. Momentum
- 56. Momentum
- 57. Momentum
- 58. Slide 57
- 59. Slide 58
- 60. Slide 57
- 61. Slide 58
- 62. Outline
- 63. Panacea
- 64. Outline
- 65. Panacea
- 66. Outline
- 67. Outline
- 68. Early Stopping
- 69. Outline
- 70. Weight Decay
- 71. Weight Decay
- 72. Weight Decay
- 73. Weight Decay
- 74. Weight Decay
- 75. Weight Decay
- 76. Outline
- 77. Dropout
- 78. Dropout
- 79. Dropout
- 80. Dropout- Intuitive Reason
- 81. Dropout- Intuitive Reason
- 82. Dropout- Intuitive Reason
- 83. Dropout- Ensemble
- 84. Dropout- Ensemble
- 85. Dropout- Ensemble
- 86. Dropout- Ensemble
- 87. Dropout- Ensemble
- 88. Practical Suggestion for Dropout
- 89. Concluding Remarks
- 1. Tips for Training Deep Neural Network
- 2. Announcement
- 3. Outline
- 4. Outline
- 5. ReLU
- 6. ReLU
- 7. Review:Backpropagation
- 8. Review:Backpropagation
- 9. Problem of Sigmoid
- 10. Vanishing Gradient Problem
- 11. Vanishing Gradient Problem
- 12. ReLU
- 13. ReLU
- 14. ReLU
- 15. ReLU
- 16. ReLU
- 17. ReLU
- 18. ReLU
- 19. ReLU
- 20. ReLU - variant
- 21. Maxout
- 22. Maxout – ReLU is special case
- 23. Maxout – ReLU is special case
- 24. Maxout - Training
- 25. Maxout - Training
- 26. Outline
- 27. Cost Function
- 28. Output Layer
- 29. Softmax
- 30. Softmax
- 31. Softmax
- 32. Softmax
- 33. Softmax
- 34. Softmax
- 35. Normalizing Input
- 36. Outline
- 37. Vanilla Gradient Descent
- 38. Vanilla Gradient Descent
- 39. Outline
- 40. Outline
- 41. Learning Rates
- 42. Learning Rates
- 43. Adagrad
- 44. Adagrad
- 45. Adagrad
- 46. Contradiction?
- 47. Intuitive Reason
- 48. Larger gradient, larger steps?
- 49. Second Derivative
- 50. More than one parameters
- 51. What to do with Adagrad?
- 52. Outline
- 53. Easy to stuck
- 54. In physical world ……
- 55. Momentum
- 56. Momentum
- 57. Momentum
- 58. Slide 57
- 59. Slide 58
- 60. Slide 57
- 61. Slide 58
- 62. Outline
- 63. Panacea
- 64. Outline
- 65. Panacea
- 66. Outline
- 67. Outline
- 68. Early Stopping
- 69. Outline
- 70. Weight Decay
- 71. Weight Decay
- 72. Weight Decay
- 73. Weight Decay
- 74. Weight Decay
- 75. Weight Decay
- 76. Outline
- 77. Dropout
- 78. Dropout
- 79. Dropout
- 80. Dropout- Intuitive Reason
- 81. Dropout- Intuitive Reason
- 82. Dropout- Intuitive Reason
- 83. Dropout- Ensemble
- 84. Dropout- Ensemble
- 85. Dropout- Ensemble
- 86. Dropout- Ensemble
- 87. Dropout- Ensemble
- 88. Practical Suggestion for Dropout
- 89. Concluding Remarks
1/89
- 速度 :
- 畫質 :
00:00/2:29:01
00:00/00:03