Skip to content

First Proposal Feedback

Jump to bottom
bzuo edited this page Dec 15, 2016 · 5 revisions

General Project Feedback

You should have all received your commented proposals. I just wanted to address some of the common points that came up and my concerns - we will also discuss in class.

  1. The data contains mixed feature types (ordinal/categorical and continuous) - think about the need for normalizing ranges and the best practices for doing this

  2. The features come from multiple platforms (clinical, protein, mutation, CNV, mRNA) and not every patient has every data type available. There is a tradeoff between including more features (which will reduce the # of patients) and including more patients (which will reduce the number of features). You do not have to use all features so think about how to build the best datasets for your experiments and a rationale for making choices. Sparse missing protein or mRNA features can also be imputed.

  3. Many of you are choosing neural networks as an approach - this is fine but please start working early. If you lack experience with NNs then it can take some time to get them working (ask Pooya or Safoora about their experiences).

  4. Try not to overextend yourselves - the project should be interesting but you are not being graded on complexity. The project is foremost an opportunity to apply what you have learned and to interpret the experiments using the fundamental concepts from class. Provide a rationale for the methods you choose, describe the hypotheses that your experiments are testing, and dissect the results carefully.

  5. Get started right away - your secondary methods use off the shelf software, and so you should be able to develop and debug a testing framework quickly and get fast results from the secondary method. You can then build on this with your primary method which will involve some algorithm implementation tasks.

Clone this wiki locally