39. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
40. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
41. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
42. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
43. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
44. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
49. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
50. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
51. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
52. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
53. Phase1
Theory
Lecture
• Definition of ML
• Issues in ML
• ML Applications
- Regression
- Binary Classification
- Multi-label Classification
• ML Algorithms
- Linear & Logistic Regression
- Activation Function: Sigmoid
- Softmax Classifier
Phase2
Theory
Lecture
• Real-world Issues in ML
- Overfitting
- Regularization
- Evaluation
• Rise and Fall of AI: The AI Winter
- Perceptrons
- Back Propagation
- Vanishing Gradients
- Activation Function: ReLU
- Restricted Boltzmann Machine
• Recent Breakthroughs
CNTK
in practice
• Environment Setup
• Logistic Regression
• Softmax Classifier
• MNIST Dataset
CNTK
in practice
• MNIST with Deep NN
- Train, Test & Evaluation
- Stochastic Gradient Descent
Homework • MNIST Multilayer Softmax Classifier Hackathon
• Dataset Preparation
• Image Preprocessing
• Binary Image Classification
Curriculum
Before & After AI Winter
54. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
55. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
56. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
57. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
58. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
59. The AI Winter Hackathon: PotatoHunter
Normal Potato v. Sweet Potato
60. The AI Winter Hackathon: PotatoHunter
@ Sinchon Dreamtop Cafe
61. The AI Winter Hackathon: PotatoHunter
@ Sinchon Dreamtop Cafe
62. The AI Winter Hackathon: PotatoHunter
@ Sinchon Dreamtop Cafe
63. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)
64. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)
65. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)
66. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)
67. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)
68. The AI Winter Hackathon: PotatoHunter
Results
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Given Example 1st Place Winner
Train Error (200 Images) Test Error (20 Images)