1. Regression vs Classification
a) What's the difference between Classification and Regression and what learning paradigms do they correlate with?
b) What changes in the training process if you use a polynomial function g(x) instead of a linear function f(x) = ax + b?
2. Support-Vector-Machines (SVM's)
a) Explain the kernel-Trick
b) How can you handle multi-class tasks with SVM's?
3. Is a standardisation necessary for the following models and why?
a) kNN
b) Random Forest
4. Large-Language-Models (LLM's)
a) What are the adventages of a RAG-Pipeline and when do you use one?
b) Name one use-case and one risk of LLM's in a production environment
5. Why are CNN's suitable for image-classification but fully connected ANN's are not?
6. A table in which to tick the suitability of different ML-Models for Classification, Regression, Clustering, Dimensionality Reduction. The models contained PCA, k-Means, Decision-Tree, LSTM, linear regression, Multilayer-perceptron...
7. Multiple-Choice tasks you should tick the true statements. Six questions but I don't really remember them.
one was "does python need setters and getters"
That was the whole exam, we got 30 minutes for it. In my opinion quite managable but I struggled a bit with the supposedly easy question 6 because I didnt remember all the posibilities of all models.
I hope this helps you!