Please do not distribute the following outside of class.
Project guidelines and deliverables can be found here.
What are effective ways to promote better ensembles of multiple models.
Related: What are ways to dynamically pick (per instance) a good model from a potential group of models.
What are effective ways of making predictions with missing features.
How can one make a meta-critic model, a model that is able to predict the performance of an other model on different instances.
Related: can one predict when a model will fail?
What are robust, effective ways of detecting outliers and out-of-distribution examples?
... (Many more are possible)
Please contact Prof. Junier if you want to brainstorm about potential project ideas.
https://archive.ics.uci.edu/ml/index.php
https://www.kaggle.com/datasets