challenge:Example Reviews

Examples of good interim feedback from 2010:

2010 Team 80
http://mode.lanl.k12.nm.us/get_interim1011.php?team_id=80

Hello all,

My name is Gary An, and I am an Associate Professor of Surgery at the University of Chicago, and also happen to be the President of the Swarm Development Group and been involved in agent-based modeling for several years. I have been asked to look at your NMSCC project and provide you with some feedback. The topic of your project is clearly very important, and as you have alluded to, one that is very highly studied. I have a couple of comments on your project update that I hope you will find useful.

1) Base on your stated goal, you are interested in finding out what factors are most affecting obesity in New Mexico, and you appear to be conducting a research survey to collect this data. As you have described it, this appears to be a statistical correlation model, and as such is a relatively standard process. It appears that you interested primarily in looking at some relations between obesity and physical activity, but appear to be waiting for your data before you go ahead.

2) So, if all you are interested in is identifying correlations, then that is all you need. You will find that some of the factors will be more correlated than others. The interesting question, however, is whether changing any of those things could be expected to make a difference. This is a much more complicated question, but also infinitely more interesting. In order to do this, you need to "construct" a mental model of what you think is going on.

3) In my opinion, you have two ways to go about this: both require you to construct some "decision process" model for people (i.e. how do they behave) and also some obesity model (i.e. how do the choices people make affect them becoming obese). There are two tacks to this: 1) you can focus on the "decision model," meaning that given a certain set of conditions (family background, location, diet), how can you change how a person decides what to do and see if that changes the number of obese people, or 2) you can "lock down" the decision model, meaning that people will behave they way they behave, but then change the inputs (like proximity to recreation, free time) and see if the number of obese people go down. I would strongly recommend that you choose one or the other, because if you try to do both at once you will have too many variables and not be able to figure out what is going on. Even if you eventually want to do both, you have to do each individually before you combine them.

4) This last thing is not as hard as it seems, because both options 1

and 2 have relatively intuitive solutions: for 1) if they eat less and exercise more then you will have less obesity and for 2) if you set a set threshold for exercise based on proximity to recreation then the closer they are to recreation then the more physical activity they will have and therefore there will be less activity. But after you have a model that is able to do both of these independently then can you look at the interactions between them (i.e "does being closer to a recreation site make you more likely to alter your decision to go to one?" or "if you have only a limited amount of time would you choose between traveling farther to exercise more vigorously or stay at home but work out less hard?). Just remember, you need to link the variables together in some sort of process.

5) Finally, when you are choosing a modeling method you want to keep in mind what sort of question you really want to answer, and intrinsic to that is whether you want do deal with space. If you think you need a discrete representation of space, then an agent-based model is a good way to go, if you don't, then a equation-based method would suffice (note that this last situation can also deal with clumps of people or things in particular "places;' you just label categories for each place; what I'm talking about with the spatially explicit is that you actually need/want to see how the things move from one area to another).

In general, a very interesting and important subject. Good luck to you!

2010 Team 22
http://mode.lanl.k12.nm.us/get_interim1011.php?team_id=22

HI my name is Eleanor Walther and I have been asked to give you feedback on your interim report. I am a

champion for Expanding Your Horizons, Project GUTS and the Challenge, being head judge of the EXPO in April. I am retired from Sandia National Laboratories, where I was the Principal Member of the Technical Staff, Emergent Threats.

Indeed the problem you pose, finding a herbal cure to cancer is a lofty goal. And if you solved this problem you would surely receive a Nobel prize in medicine. I'm not sure why you think finding a herbal cure will save the patient money. There are existing chemotherapies that are very expensive that have herbal origins.

Cancer is not a single disease. Even within a cancer type there are many different diseases. If you want to continue with this project, I suggest that you pick one subtype of cancer. There have been past SCC projects on cancer topics so I would recommend that you go through the SCC archives and look at past projects on cancer to get an idea of how you might scope this project down. That will help you focus your research. Wikipedia and google are not research references. Go to the Resources tab on the Challenge wepage and scroll down and take a look at the research section to get an idea of how to start your research.

After you have done these things I think you will be able to refocus and restate your project proposal so that it is doable in the Challenge year time frame.

Good luck and congratulations for the hard work you have done so far.

2010 Team 65
http://mode.lanl.k12.nm.us/get_interim1011.php?team_id=65

Hello Team 65 -

My name is Roger Critchlow. I am a software developer in Santa Fe. I have worked on computers for 30 years, developing many different kinds of computer systems, including population biology models, computational chemistry, repair scheduling for airlines, operating systems, and software development tools. I'm currently working on distributed file systems. I have also volunteered for the Supercomputing Challenge for many years, and they have asked me to review your interim report.

Your report confused me. Your summary speaks almost entirely of transmission lines and the basic electrical engineering involved. But all your references appear to be about applications of neural networks to prediction, optimization, and simulation problems. I think I understand how those two topics might come together in your project, but you will need to make it clearer when you present it at the evaluations next month.

I think you should find some references about electrical engineering, too, because some of what you're saying isn't completely sensible. For instance, you state that: "To reduce the resistance on these power lines you also must increase the voltage." I don't think you can mean "resistance" in the technical, electrical engineering sense, because resistance is a static property of conductors that doesn't usually change, even when you increase voltage. If you're going to use a technical word like "voltage" in a sentence, then any related technical words in the same sentence -- like "resistance", "current", or "power" -- should only be used within the same technical context.

While googling one of your references, I stumbled on to this paper by two Korean researchers:

http://www.intellcontrol.com/files/kc/NN3- Application%20of%20neural%20network%20controller%20for%20maximum%20power%20extractio n%20of%20a%20grid-connected%20wind%20turbine%20system.pdf

It's what I think you're aiming for, in the long run. They train the network to control the turbine blade pitch to generate the most power for wind.

I think you've picked a very interesting project and I hope to see an excellent final project from you in April,