berkeley-stat159 / project-kappa

BSD 3-Clause "New" or "Revised" License
0 stars 7 forks source link

feedback #39

Closed jarrodmillman closed 8 years ago

jarrodmillman commented 8 years ago

One thing that stood out: aiming to outperform a previous analysis is a very strong claim - I would reorient the goal to focus more on evaluating the different machine-learning methods rather than focus on outperforming the original analysis.

Scope of the project seems very large - Need to explicitly express what they are trying to accomplish with machine learning (supervised learning [classification / prediction], unsupervised). "Machine learning the hell out of" the data is not a useful statement - different machine learning approaches are designed to accomplish different things. Need to spend more time explicitly formulating the problem to be solved.

Computer performance was ID'ed as a problem - this surprises me, may want to start with subsets of data for exploratory analysis and scale up as necessary.

Presentation didn't contain any preliminary analysis results - really need to get moving here. Perhaps start with the analysis from class.

You still have a lot of work to do, but there is still time. However, you will need to start focusing and making consistent progress. If you continue waiting to start, I am worried you won't have enough time to make serious progress. Please take advantage of our office hours.