1. select any of the datasets you can find on the UCI Machine Learning Repositoryhttps://archive.ics.uci.edu/ml/datasets.php 2. write some small Python script to construct a neural net and apply it to...

1. select any of the datasets you can find on the UCI Machine Learning


Repositoryhttps://archive.ics.uci.edu/ml/datasets.php



2. write some small Python script to construct a neural net


and apply it to the dataset


3. report your findings: e.g. what architecture did you use, what parameter


combinations did you try, what were the respective results.


Obviously you can use the sample code on the lecture website



http://mitloehner.com/lehre/ai/



but try to make it your own: add more layers and observe the effects,


use different activation functions, embedding dimensions, batch size,
optimizer, ...


maybe even try to work on one of the datasets with non-numeric inputs.


Do NOT use one of the datasets or the exact code we covered in the lecture


(Iris, MNIST, Pima, ..)




Send me an email with your code, the results, and a short text summarizing


your findings.