Choose a dataset from the UC Irvine Machine Learning Repository with at least 5000 instances and 20 attributes for classification or regression. Compare how the different approaches seen in class perform on this dataset to predict accurately the classes or the values of the unlabeled data. You should determine what are the best hyper-parameters for each approach you are using. You could use any Python libraries.
You will form groups of 3 or 4.
- Presentation of the research questions and the chosen methods to tackle them
- Relevant Literature Review
- Presentation of the results and discussion
- Conclusion/future work
We used tensorflow-cpu
, since it works universally on all systems, and for some reason it runs much much faster than tensorflow-gpu
for our testing.