To work in the machine learning research lab, it is important to be able to read and understand machine learning research papers. It is also important to have excellent written communication skills. Therefore I need to judge your reading/writing skills and the quality of your scientific comprehension/ideas. Please take some time to choose a publication that is interesting to you, and for which you think there would be an interesting future research project/publication. Then write me an email, in your own words (not copying from the publication), with
- Background: what is the problem setting and data? e.g., for regression the data is a n x p input/feature matrix, an n-vector of outputs/labels, problem is learning a function which takes a p-dimensional input/feature vector and returns a real-valued output/prediction.
- Previous work: what are the existing approaches for that problem, and what are their drawbacks that motivate a new algorithm? e.g., for regression there are linear models, neural networks, boosting, etc.
- Novelty: what are the new ideas presented in the paper? Are they theoretical or empirical, or both? e.g., the paper could use existing models/data and present a new proof about the optimality/speed of an existing algorithm, or it could use existing algorithms/data with a new neural network model architecture, or it could present new benchmark data sets for comparing various existing algorithms/models.
- Results: what comparisons were done to show that the new idea is interesting/useful in theory and/or in practice? e.g., you could compare the test accuracy and computation time for different regression algorithms on various data sets.
- Future work: what are some of your ideas for new research papers that could be written as a follow up to this one? Justify why these ideas are sufficiently novel that they warrant a new paper describing them.
Last updated 30 Dec 2020.