Deep Learning in five steps.

Lecture 1

• 1.1. Preliminaries
• 1.2. My Works
• 1.3. Some historical remarks
• 1.4. Deep Learning (DL)
  • 1.4.1. DL in AI
  • 1.4.2. Machine Learning (ML)
  • 1.4.3. Deep learning
  • 1.4.4 What types of the problem DL solves
• 1.5. Linear regression problem
  • 1.5.1 Some theory
  • 1.5.2. Simple implementation
• 1.6. Keras for linear regression

Lecture 2

• 2.1 Generalities of Keras
• 2.2. Nonlinear regression - Neural Networks
• 2.3. Universal approximation theorem
• 2.4. Propagation of signal and back-propagation of error
• 2.5. The general strategies for Gradient Descent Optimization
• 2.6. The popular gradient descent algorithms
  • 2.6.1. Gradient Descent with the momentum
  • 2.6.2. AdaGrad
  • 2.6.3. RMSProp
  • 2.6.4. Adaptive Moment Estimation (Adam)
• 2.7. Predicting Boston Houce Prices - a nonlinear regression problem
• 2.8. Vanishing gradient problem

Lecture 3

• 3.1. Cross-Entropy for two clsses - Binary Cross-Entropy
• 3.2. Binary clssification: IMDB data
• 3.3. Multiple independent attributes
• 3.4. More than two classes mutually exclusive
• 3.5. MNIST a “Hello Word of Deep Learning”
• 3.6. Basics of the Convolutional Neural Network (CNN)
• 3.7. General remarks on ConvNet
• 3.8. ConvNet in Keras
• 3.9. DropOut Layer
• 3.10. Batch normalization layer
• 3.11. MNIST from ConvNet

Lecture 4

• 4.1 Getting Data From Kaggle
• 4.2. Preprocessing and data augmentation
  • 4.2.1. Defining Model
  • 4.2.2 Data Processing – Generator
  • 4.2.3 Data Augmentation

Lecture 5

• 5.1 Pretrained models