10. Workflow in Supervised Learning
10
1. prepare the data
2. extract the features
3. select the model
4. select the parameters
5. train the model
6. evaluate the results
1 2
3/4/5
6
1 2
11. Workflow in Supervised Learning
11
1. prepare the data
2. extract the features
3. select the model
4. select the parameters
5. train the model
6. evaluate the results
1 2
3/4/5
6
1 2
which steps could
be automated?
12. Finding a Deep Learning Solution
12
• Selecting and Designing Network Architecture
- Multi-Layer Perceptron (MLP)
- Convolutional Neural Network (CNN)
• Selecting and Optimizing Hyperparameters
- number of layers and neurons
- objective function, activation function, optimization algorithm
- batch, epoch, learning rate, momentum, dropout, ...
• Using State-of-the-art Model and Pre-trained Weights
13. Finding a Deep Learning Solution
13
• Selecting and Designing Network Architecture
- Multi-Layer Perceptron (MLP)
- Convolutional Neural Network (CNN)
• Selecting and Optimizing Hyperparameters
- number of layers and neurons
- objective function, activation function, optimization algorithm
- batch, epoch, learning rate, momentum, dropout, ...
• Using State-of-the-art Model and Pre-trained Weights
which steps could
be automated?