NYU K12 STEM Education: Machine Learning (Day 9)

nyu-header

Final Project

  • Task: CIFAR10 classification.
  • Simplification: No validation. You can directly tune your network based on the test set accuracy.
  • Friday: Present your model performance.
  • Each team should present for 12-15 minutes.

Presentation Template

  • Slide 1: Title and introduction
  • Slide 2: Introduce your project: dataset, labels, etc…
  • Slide 3: Your network architecture
  • Slide 4: Other hyper-parameters, e.g.
    • optimizer, learning-rate
    • batch size, epochs
    • data augmentation, transformation
  • Slide 5: Model performance on training set (loss) and test set
    • Training loss
    • Test accuracy total
    • Test accuracy for each class
  • Slide 6: Challenges and how you resolve them.
  • Slide 7: Conclusion

Starter Notebook

  1. CIFAR10