This project contains all the code developed for the courseworks of the Data Mining & Machine Learning class of Heriot Watt University as a part of the Master program.

This represents the work of 4 master students:

• Dimitri Accad (MSc in Artificial intelligence)
• Antoine Auzimour (MSc in Artificial intelligence)
• Clarence Deltel (MSc in Artificial intelligence)
• Thomas Di Martino (MSc in Artificial intelligence with Speech & Multimodal Interaction)

Split in two courseworks, the first one was focusing on bayesian learning as well as unsupervised learning while the second one was more about decision tress/ensembling techniques as well as Deep Neural Networks (MLPs & CNNs).

For this assessment we have worked with the a dataset sample from Stallkamp et al's German Street Sign Recognition Benchmark which consists of:

• 10 classes
• 12660 images (converted in grey-scale with pixel values ranging from 0 to 255 abd rescaled to 48*48px)

The 10 labels to predict were:

0. speed limit 60
1. speed limit 80
2. speed limit 80 lifted
3. right of way at crossing
4. right of way in general
5. give way
6. stop
7. no speed limit general
8. turn right down
9. turn left down