Course info

Subjects covered

Aim

Introduction to machine learning from a physicist's perspective, with the aim to understand how it works and less emphasis on tricks or parameter optimization

Subjects

(order may change)

1. Deep learning (from scratch in numpy), Higher level implementations (tensorflow, sklearn, keras)
2. Convolutional neural networks
3. Image segmentation
4. Pre-trained models, big models
5. Data augmentation, auto-encoders
6. Diffusion models
7. Genetic algorithm
8. Q learning, reinforcement learning
9. LSTM networks
10. Textual data
11. Attention modules
12. Unsupervised learning

Requirements

• Classwork solution with hard deadlines 70%
• 70% participation on practice courses
• Two projects one fixed one of your own choice
• OR Test at the end of the semester and a project

Evaluation

• Practice solutions (pair): occasional varying number of points (extra points!)
• Projects: 100 points each
• Test: 40 points, those who fail the test cannot get a grade
• Marks (all points summed up):
  1. -109
  2. 110-139
  3. 140-169
  4. 170-199
  5. 200-

Consultation

During the classes or on demand by email.