Autoplay
Autocomplete
Previous Lesson
Complete and Continue
312. Build a Neural Network Framework
1. Build a Testing Data Set
Get started
How to get what you want out of the course
A Roadmap
Exercise 1. Create a set of 2x2 pixel toy example images. (4:24)
Diff-viewing tools
Exercise 2. Write training and evaluation set generators. (4:34)
Exercise 3. Make a runner that loads the data (3:09)
2. Build an Outline of a Neural Network Framework
Why Autoencoders?
Exercise 4. Build an ANN stub. (3:54)
Exercise 5. Add examples to training and evaluation. (2:40)
Exercise 6. Add expected data range and normalize(). (5:47)
Exercise 7. Add a denormalize() function. (3:15)
3. Build a Fully Connected Linear Layer
Exercise 8. Create a Layer object. (4:55)
Exercise 9. Add linear forward propagation. (3:47)
4. Build a Multi-layer Nonlinear Network
Exercise 10. Add a nonlinear activation function. (4:21)
Exercise 11. Add a single hidden layer. (2:44)
Exercise 12. Expand to several hidden layers. (0:55)
5. Find and Report the Error
Exercise 13. Write error reporting. (8:53)
Exercise 14. Write squared error function for autoencoder output. (3:17)
Exercise 15. Add the error function in to the final stage. (3:09)
Exercise 16. Add an absolute value error function. (2:01)
6. Integrate with Visualization
Exercise 17. Incorporate visualization. (3:59)
Exercise 18. Write forward_prop_to_layer() and forward_prop_from_layer(). (7:34)
7. Add Backpropagation
Exercise 19. Add backpropagation scaffolding. (6:16)
Exercise 20. Add tanh() derivative (3:10)
Exercise 21. Add logistic and ReLU activation functions (2:18)
Exercise 22. Adjust weights by the full gradient. (3:21)
Exercise 23. Write the learning rule. (1:52)
8. Build an Advanced Testing Set and Refine the Framework
Exercise 24. Create a set of 3x3 pixel examples. (2:06)
Exercise 25. Run the 3x3 testing set on the framework (2:06)
9. Build a Demonstration Data Set and Play with the Framework
Exercise 26. Run framework on Nordic runes data set (6:25)
Exercise 27. Adjust the number of nodes (3:32)
Exercise 28. Adjust the number of layers. (5:15)
Wrap Up
Exercise 22. Adjust weights by the full gradient.
Lesson content locked
If you're already enrolled,
you'll need to login
.
Enroll in Course to Unlock