San Diego (Winer ANS)

Been a busy few weeks.  Last week I successfully defended my master thesis in a very awkward presentation.  Then on Sunday I flew out to San Diego for the Winter ANS meeting.  I think I like San Diego.  Good city.  Easy to get around in - only city where buses were actually on time.  I also really liked the Gas Lamp district; it is a thousand times better than Knoxville.  Maybe even the 'burgh - but only because it was condensed  (think of combining the bars of Shadyside with Oakland and Southside).  The presentation went very well (expect for not knowing my left and rights, but those are details). And now I fly back to Knoxville, do a Data Mining presentation tomorrow at 1, and then on Wednesday it is back to the farm.

Master Thesis

Summary: tldr;

Presentation on Thursday, come see it, I'll bring candy.  Then San Diego on Sunday-Thursday.  Then The Farm Wednesday for Thanksgiving.

I got to see RR today!

Sandy

Some photos from David of the Farm after Sandy:

Good attachment to the purloins

The wire is still intact, the barbed staples pulled from the post.

Energy Depostion (Co-60)

One of the radiation portal monitor specifications is to be insensitive to gamma rays due to a large natural gamma ray background (medical isotopes, cat litter (potassium), bananas, etc). We are then investigating how much of the energy (from Co-60) is deposited in a film compared to the amount of energy deposited by the charge products of Li-6(n,t)He.  Using geant4 in parallel (shooting a few billion particles, takes about 2 hours on 80 cores) I grabbed the energy deposition for various film thickness, ranging from 15 microns to 1 cm. 

Geant4 Simulations

When compared to the measurements I do fairly well.  The measurements aren't really energy deposition but really pulse height spectra which includes optical photon generation / pulse height deficit, light collection, and PMT effects so I didn't make an effort to find a scaling factor from the simulated to the measurements because it would inevitably involve a fudge factor.  By just plotting the average I got around this (integration sweeps all sorts of stuff under the rug) and can then argue that because I got the trend correct for what I am intrested in, my simulation is good.

Actually, all that it tells me is that it isn't wrong, but doesn't prove that it is correct.

Parallel Geant4!

After a long and frustrating process I finally got Geant4 to run in parallel!  Who knew setting up this stuff could be so hard?  I liked it better a Fermi Lab / CMU where all the environment was taken care for you.

Now, to the forefront of nuclear science!  Researchers at UTK / ORNL want to create ultra cold neutrons, slow them down to ~10-30 K, and "drop" them down a 100 foot tube to look for flavor oscillations of the quarks in neutron - anti-neutron oscillations.  It's a really nifty idea!  I didn't know that neutron's could oscillate between anti-neutrons, but it makes sense; I mean, why not?

Expression Trees / Networkx

Our next lab in machine learning is symbolic regression (fitting a function) using genetic programing.  It's a neat (though a little bit contrived) example of genetic programing.  Our programs are expression trees which we mutate according to a fitness function (the least square error).

Simple Expression Tree

Our actual function to fit just looks to be a sum of sine's, so it would be really cool if I could find an expression tree that would just take the Fourier transform, pick out the principle components, and return as sum as sines.  I have been using the networkx (published by LANL) and so far it has been really nice, but the trees aren't recursively defined so it takes a little bit of thinking to evaluate (rather than a simple post order traversal). Plus (the reason why I choose it) is the interface with Graphviz, which makes it quite easy to visualize trees.  The windows interface is a little clunky (use dotty to draw) but it worked great on the Mac.

Now if only my real research was going so well . . .

Frequency of Names

A while back I got into playing with the cenus data on baby names in order to show the odds between picking a girls name that ends in 'ie' vs 'y' - I always thought that 'ie' would be the more probable choice, but that could be because of dating a girl whose name ended in 'ie'.

 I downloaded the data set from  the SSA (http://www.ssa.gov/oact/babynames/limits.html), and parsed it for the total number of females born in a year, and the number of names that in the the sufix 'ie' or 'y'. Well, in 2011 11% of the girls named ended in 'y', while only 5% ended in 'ie'.  So if you were guessing on the spelling go with 'y'!

 I also looked a the occurrence of biblical names by finding a data set of about 2,000 names that were mentioned in the bible.  I didn't do anything clever with different spellings, just a straight string comparison.  The decline up to the 20's make sense, but then the divergence between the male and female names don't make sense.  This could be because of spelling.  I also find the increase in biblical names in the 1940's to 80's very curious.  The greatest generation falling back on their roots?  Or simply past the 80's the hippies started to have kids named Sunshine?

 Along a similar vein I tried to look if parents were more likely to name their child after a president. I tried to normalize out the effect of names become more popular by using a normalizing the amount of children born during a presidents term by the number of children born in the last 10 years of that name, but I don't think I was successful, simply because the strangest presidents and vice presidents corresponding to the peaks (people name their kids after Dick Cheney?).  The only peak I think if physical is the 1960's, for JFK.  I think the main problem is that most presidents have pretty non-nondescript names, which have a very fluid popularity.

Legend should read POTUS, !PTUS

Rebecca Ruth Lee

I 'm an Uncle!

Katers and Rebecca Ruth (RailRoad)

In all seriousness, congratulations to Katie and Mike. I look forward to meeting the bugger over Thanksgiving.

Movie Prefrences and Fruit Update

Apples! They grow, yo!

Thanks to David for diligently recording

I find it very fascinating looking at how many apples have been picked.  2,500 baskets seems like a pretty small number though.  But hey, at least we finally picked some pears! 

Confusing little plot where the hotter colors indicate a higher ranking.  I was hoping that by looking at the data I could find an attribute (zip code, age, gender) that would be a good first indicator to split on. 

I seem to have a small off by one error, or maybe pmeshcolor isn't the right plot to use / doesn't do what I think it does. But some bits of it do seem to make sense; Males are more likely to watch comedy, while females go for the drama. I guess I need to center the grid lines, but I couldn't figure out how to.

Zip Codes

As part of my data mining class we are trying to write a decision tree to suggest movies, based off of the MovieLens data.  It's a neat little project (and fun - I love recursion and trees).  I got to looking at Zipcodes, and came across this little gem.

http://eagereyes.org/zipscribble-maps/united-states

In another news, coming back from Sectionals Fong killed the radio in the Red Rocket by playing country music.  No lie, after about 30s of playing country music the Red Rocket Radio stopped working.  Now I am just going to put a boom-box in the passenger seat - ghetto yo!

Going to my first night football game, Go Vols!

Electron Energy Depostion in Water

In 1982 Turner published a paper on the energy deposition in water, and I have finally managed to recreate the figure in GEANT4.  I think that the G4_Water material has better properties than that in the 80's, which explain why my results have a slightly different finer structure.

Single collision energy loss spectra.  Insert is Turner data.

The trick was grabbing the kinetic energy after the first inelastic collision, which is not the final kinetic energy reported at the end of a trajectory.  I got this result the brute force way - verbose stepping, and then piping the output into sed to search for the first track until the a process name (e-_G4DNAIonisation), pipe that output into grep to grab the appropriate line, then into awk to grab the 5th column.  Not pretty, but I'll take it.

I'm really enjoying this stage of the project; it is just enough of a challenge to keep me awake at night working on it without becoming discouraged.

Linear No Threshold Model - Applied to Fukushima

This WSJ article provides a nice little bit about the doses and cancer risk from Fukushima. I'm not going to summarize it because y'all should read it.  In radiation dose/ risk there is a continuing debate over the effects of low doses of radiation.  There is some who believe that low doses are actually good for you (make the body's repair mechanisms more robust for larger doses*) and another group who believes the risk of getting cancer is linear with the dose. At every ANS conference there is always some group trying to put this question to rest, but the results are always sketchy - this is all a very complicated problem because even those some groups get different doses (i.e. next to a uranium mine, atop of a mountain or living in a basement with radon) the statistics are not there and it is difficult to compare between lifestyles and populations.

What did the stop light say to the car?

Qualified!

I'll summarize the quals in these images:

Favorite Question: A miracle in materials science allows for lead to be used as a scintillator . . .

Least Favorite Question: Consider the following cross sections provided for three group diffusion . . .

Results:

Baring delays, takeoff is scheduled for May, 2013!