Chapter 1. What is deep learning?
Figure 1.1. Artificial intelligence, machine learning, and deep learning
Figure 1.2. Machine learning: a new programming paradigm
Figure 1.3. Some sample data
Figure 1.4. Coordinate change
Figure 1.5. A deep neural network for digit classification
Figure 1.6. Deep representations learned by a digit-classification model
Figure 1.7. A neural network is parameterized by its weights.
Figure 1.8. A loss function measures the quality of the network’s output.
Figure 1.9. The loss score is used as a feedback signal to adjust the weights.
Figure 1.10. A decision boundary
Figure 1.11. A decision tree: the parameters that are learned are the questions about the data. A question could be, for instance, “Is coefficient 2 in the data greater than 3.5?”
Chapter 2. Before we begin: the mathematical building blocks of neural networks
Figure 2.1. MNIST sample digits
Figure 2.2. The fifth sample in our dataset
Figure 2.3. A 3D timeseries data tensor
Figure 2.4. A 4D image data tensor (channels-first convention)
Figure 2.5. Matrix dot-product box diagram
Figure 2.6. A point in a 2D space
Figure 2.7. A point in a 2D space pictured as an arrow
Figure 2.8. Geometric interpretation of the sum of two vectors
Figure 2.9. Uncrumpling a complicated manifold of data
Figure 2.10. Derivative of f in p
Figure 2.11. SGD down a 1D loss curve (one learnable parameter)
Figure 2.12. Gradient descent down a 2D loss surface (two learnable parameters)