The Ingenious Gentleman Dan Lakeland of LaFayette

Well, for those of you interested in such things, here's the story so far...

The math class that I took without the required prerequisites turned out to be a great way to learn the prerequisites. it wasn't very applications oriented, but it was an area of mathematics that has nice applications and therefore I'm pretty happy to have taken it. However, it did take up a lot of time learning about Sobolev spaces, probability measures on function spaces, weak solutions to PDEs, a-priori estimates, martingales, Ito and Stratonovich integration, Ito's formula, generalizations to Hilbert space valued martingales... basically a lot of material.

My multiscale mechanics class fit nicely with the math class. I am working on a side project related to multiscale modeling in the context of Schwartz distribution theory. It's a variation on the Fermi-Pasta-Ulam problem as written up in American Scientist this month.

As for research, I am still struggling to write the proper equations for conservation of energy. the issue is that I can not make many of the simplifying assumptions normally made, for example I have a compressible fluid and temperature dependence of specific volume and pressure. I'm planning to talk with Maartin wednesday on this.

Nevertheless I have some approximately correct equations, and i have verified that they behave in the proper way, however in addition to being complicated, the equations are stiff. The speed of sound in water is very fast, so I need to take very small timesteps. Once I have my equations properly derived, I plan to work on simplifying by splitting the fast and slow components, eliminating terms that are negligible compared to other approximations I am making, and either working in a purely numerical scheme, or being smarter about my use of computer algebra to get a better timestepping method.

In the summer I'm working on these projects, and talking with my experimentalist advisor about experimental methods so that we can validate my basic model for publication hopefully soon after the summer.

One hears a lot about cause and cure for economic ailments these days. My take on this is based on a Milton Friedmanish view of the effect of the (broader) money supply.

Basically, houses became a source of money for those families willing to participate in the bubble, so long as prices were increasing, interest rates were low, and lenders were willing to lend against equity. However, when the bubble burst, this form of extended money supply collapsed and took the companies who had financed this inflation with it. When this portion of the money supply dwindled those goods that had inflated in price with it, such as perhaps rents, automobiles, food, airline travel, cell phone services, consumer electronics, etc were no longer priced at viable levels. The massive reduction in demand for those types of goods has rippled into companies across the board. Anyone who bought capital equipment in the last 3 or 4 years basically overpaid relative to the current money supply. This LOOKS like a loss in monetary terms (only economists think in real terms unfortunately, everyone else pretty much trusts accountant type calculations).

The reason I believe this is the main issue in the whole phenomenon is that I lived in an area that saw this effect from way back. The San Francisco bay area was a center for the dot com boom. If you wanted to buy a house in the east bay in 1997 you could probably get one for 300k, by 1999 they were 700k, by 2005 they were 900k to 1.5M. The dot com effect was localized to a few areas around the country, but it made a LOT of money for finance companies. When companies IPOed at $15 and went to $150 overnight the company got $15/share, but the finance companies working the deals got the other $135. One of the major reasons Google is so successful is that they refused to dance that dance, and pre-auctioned their shares at fair market value, resulting in a rich tech company, not a rich investment bank.

This manufacturing of money seems to have directly led to an oversupply of easy loans, which combined with artificially low interest rates after the dot com collapse to give us a housing boom. The housing boom affected everyone since there was no need for specialized technology companies which tend to localize to regions of high technology concentration.

If you were sitting on the sidelines saying "These people are crazy" and living in tiny apartments or with your family members during this whole 10 year debacle you are now helping to pick up the check for those who flaunted reality. The way in which this works is that the government is going to spend your money for you now ($900 billion or so, about 7% of GDP?) and take it from you later. In the mean time, they're hoping that this increased demand for government purchased goods, combined with an infusion of ready cash into lending will increase the general prices back to levels that are similar to what people have gotten used to and people will stop holding cash in expectation that they'll be able to buy more stuff in the future if they just wait for prices to come down.

If I were in charge, I suspect I would prefer to do something along the lines of the following:

1) Slash tax complexity in a major way (the idea here being to undo a lot of tax-incentivized behavior and make it easier for companies to plan economic activity)

2) Drastically reduce income taxes, and create a wealth (property) tax. Income taxes create disincentives for work, wealth taxes create disincentives for hoarding money in non-productive ways. A property/wealth tax would be based on stocks, bonds, cash savings accounts, automobiles, homes, other real property, business equipment, and other large ticket items that are easy to identify without intrusive jack booted thugs rifling around under your mattress.

3) make it easier for GM, chrysler, and other big companies that have been hurting to go out of business. The last thing we need is wealth being wasted by companies that produce products that consumers don't really want.

4) Simplify tax incentivized savings: Create a SINGLE account that anyone can open which takes the place of 401k, 403b, IRA, RothIRA, etc and which people can put up to say 25% of average income into pre-tax, and use for human capital development type purposes (education, health care, housing, investment for retirement, childcare, etc).

But I'm not king, and I doubt anyone much is listening.

I just gave a "15 minute" presentation to my stochastic partial differential equations class.... My suggestion to all those giving timed presentations is:

LOOK AT YOUR WATCH BEFORE THE START OF THE PRESENTATION.

The people ahead of me went over their time, and then when I looked down it seemed like I'd only been talking for a minute or two and it already said my time was up... OMG PANIC... look totally disorganized, etc...

The real issue is that I started 10 minutes later than when I thought I'd started... so my 15 minute presentation became a 5 minute presentation by my watch...

other than that, it went really well. the professor stayed after class to ask questions about the derivation and was clearly interested in the application I was describing.

Mathematicians are people who are interested by equations.

Scientists are people who are interested by the phenomena described by equations.

Engineers are people who are interested in the money saved by describing phenomena by equations.

Economists are people who think that equations about money are interesting.

Funny thing happened today. I went to my advisor's office to talk about the stochastic PDE class. During the class he looked up several things on the internet. At one point he says "hey that's your livejournal!". apparently he had googled some terms that were similar to ones from this post about the SPDE course and it came up about 3 or 4 on the google search 8-P.

I thought it was humorous, but it reminded me that everything can be googled and one should not post anything too sensitive in your livejournal...

In other news he had is 2 week old baby with him. SOO cute.

It's a graduate "Topics in Probability" course on stochastic partial differential equations...

It's challenging because I don't really have even close to the background that I should have.

1) Never taken measure theory and formal probability (took mathematical statistics, combinatorics, lots of experience in applied stats and data analysis, experimental design, etc)

2) Never taken a stochastic processes course (martingales, brownian motion, continuous processes etc, but have some familiarity with discrete markov chains)

3) Never taken a stochastic ODE course (ito integrals, stratanovich integrals, etc)

4) Taught myself most of what I know about Hilbert/Sobolev spaces and PDEs by reading books.

5) Been since 1994 when I last took an ODE course (ODEs oriented to engineers with computer component using Mathematica).

I'm skeptical whether I can really get this material, but at the least it's giving me a good reason to learn the prerequisites (ie. stochastic processes, measure theory, Lebesgue integration).

When I graduated from ISU you could get a math degree without taking Real Analysis. They changed that... as they should...

So if you don't hear from me for several months it's probably because I've got my head buried deep in some probability theory notes somewhere... or reading about multiscale mechanics... my other research oriented course... ummmm yay, I guess...

Today at 7-9 PM is my last final, in dynamics of structures (vibration theory). The test covers a lot of material, and is very conceptual oriented (probably little in the way of calculation, much in the way of derivations and explanations, qualitative answers etc).

I did very well on the midterm (got 100%) so I'm hoping to do well on the final. I like tests that focus on concepts. I just wish it wasn't scheduled so late in the day!

I didn't get a good picture, but the B2 bomber was very visible directly above my house low enough that you could probably see the pilot if it were inverted (which it wasn't).

Here's a wikipedia image:

The decline of western civilization?

My mom has a bunch of little games where pegs go in holes in a board, and you have to move them around via some rules to achieve a goal. One game goes like this:

There are 9 holes in a line. 4 red pegs go on one side, then a space, then 4 blue pegs on the other.

The goal is, by moving the pegs monotonically towards the other side, switch the pegs so that red is all on the other side and vice versa. Legal moves are a move from an adjacent position into the space position, and a jump over a peg into the space position.

I bet my mom I could quickly write a simple computer program to solve the problem, which I did last night. Challenge to you all: write a shorter computer program than I did. 7 lines of Prolog code.

Code in the comments below (after a short break).

Luca Turin talks about his theory of smell, and why he's able to make a bundle of cash by finding alternative molecules for perfumers using computational tools.

For those friends and family who are not quite sure whether we're near the Sylmar fire that's raging in LA, you can rest easy. We're about 20 miles south east of where the fire is, and it's heading west so away from us if I understand correctly. But we are getting a lot of smoke in the whole area.

Books that have been influential (see Paul's list for explanation). Building the list first, and tricking it out with links and explanations later:

• The Feynman Lectures on Physics, esp. vol 1 and 2: Not just for their content, but also for their dramatically radical teaching style. Excessive for their original purpose (an undergrad course at Caltech!), but perfect for the student of physics who doesn't have to "get a grade" but wants to know how it really works.
• The Language Instinct: Why everything you were ever taught about languages by your high school teachers is wrong, and what this means about the brain. Should be required reading for all English teachers at the high school level.
• Zen Mind Beginner's Mind: Shunryu Suzuki roshi (teacher) introduces you to the practice of Buddhism in the Soto zen tradition. This is a book you will give away again and again.
• Practical Applied Mathematics: Most books on mathematics focus on solving problems and proving theorems about solutions. This book actually focuses on how to build a model. Most of the mathematics would typically be understandable by a junior or senior undergrad, but courses rarely talk about building models. This also emphasizes the importance of nondimensional models.
• Bayesian Data Analysis: Andrew Gelman shows you how statistics really ought to be done. If you work with data, you need this book.
• Modern Applied Statistics with S-Plus: Venables and Ripley give you a major toolkit for data analysis in R (the GNU version of the S language). If you work with data, you need this book (updated version includes R specific stuff).
• Why Buildings Stand Up: Mario Salvadori explains the basic concepts of structural and architectural design, with a lot of nice examples and sketches.
• Why Buildings Fall Down: Salvadori again, on forensic analysis of failures. Fantastic book, and includes a section on the twin towers, written by a second author (Salvadori is dead now).
• On Lisp: Paul Graham shows you how to think about programming and metaprogramming in the most metaprogrammable language that has ever been invented.
• Scheme and the Art of Programming: This was the beginning programming book used at ISU to introduce CS majors. It's a good introduction to very high level programming concepts. Much better than your typical "C++ how to program" or whatever. It probably isn't as good as SICP but I haven't actually read SICP.
• Bicycling Science: A review of the science and mechanical engineering involved in the bicycle. A very nice overview of some concepts in mechanical engineering, as well as a great inspiration for bicycle riders.
• fill in more...

Just finished a strong application for an NSF Graduate Research Fellowship in Civil Engineering. One way to get back the taxes taken from previous paychecks :-)

With any luck, things will go through, and I will be able to afford to finish my PhD and produce a kick-ass risk model for soil liquefaction, while at the same time advancing multi-scale mechanics of materials.

now I just need some of those books from Amazon to arrive...

For friends planning to be in the bay area around thanksgiving, I'm going to be there wednesday-saturday the week of thanksgiving. Give me a call!

UPDATE: NSF says transcripts were received correctly so i can stop stressing out!

Assume my model for elections is correct, it shows that as the number of people who vote decreases, the probability of affecting the outcome increases. Similarly, as the distribution of the voting probably p shifts away from 1/2 the probability decreases.

There are non-negligible costs to voting, including time spent researching, watching debates, propagandizing for your favored side, standing in the rain, and potentially harassment at some locations. I hypothesize that at the margin, the average cost to society equals the expected value of the additional vote.

Gelman argues that voting is rational if you believe that the expected value per person of your choice is positive, then the overall EV to society is 300e6 * ev/person. I further hypothesize that although this argument still holds, since everyone will continue to do their songs and dances, the expected value of the net benefit is negative... all that campaigning and distracting and soforth has a nontrivial cost.

Perhaps it's all really a consequence of the 2nd law of thermodynamics.

The presidential election is a complicated process involving electoral votes and soforth. For an analysis of whether your vote matters to the presidential election, you might be interested in Andrew Gelman's analysis.

A simpler model for a simple binary majority vote, such as on a controversial ballot initiative (Prop 8 for example) goes as follows:

California has about 36.5e6 people, assume about half +- 5e6 will vote on the initiative. (note, the results are not strongly sensitive to this number. I simulated 1/4 will vote and got a result that differs by a factor of less than 2)

Assume that the probability of a person voting yes on prop 8 is 1/2 +- .01... In other words, it is a tightly contested issue.

We simulate the probability of a tie in overall voting that could be broken by your vote.

1000 times we do the following: choose a probability from the normal distribution with mean 1/2 and stddev = .01. Then for each probability value, we simulate 10 elections, where n is chosen at random according to the estimate of the number of people who will vote, and k (the number of yes votes) is required to be a tie, the binomial distribution is used to simulate the election.

R code follows:

sims<-replicate(1000,
{x <- round(rnorm(10,36.5e6/2,5e6));
dbinom(round(x/2),x,rnorm(1,1/2,.01))})
dim(sims)<-c(10000,1)
sims <- sims[!is.na(sims)];

mean(sims);

[1] 2.547123e-06

The expected probability of your vote mattering on a tightly contested ballot election (the "best" case) is about 2.5 out of a million. If we assume 1/4 of the population will vote +- 2.5 million we get 3.6e-6 instead. In other words, it doesn't matter how many people vote exactly, it is still only around a 1 in a million chance that your vote will break the tie changing the course of the election.

(note, the simulated values are not terribly stable with this small number of simulations... but they are always between about 2 and 5 x 10^-6, heck call it 10e-6).

With our generous estimate of 10e-6 as the probability of affecting a simple election like this, what is the probability that you will NEVER affect an election. Suppose you live to vote in 100 elections. dbinom(0,100,1e-5) = .999

In other words, there is a 1/10000 chance that your vote will affect a statewide simple-majority election in your long-extended lifetime. My guess is volunteering even once to man the phones to campaign for your issue does far more than actually voting.

There are several very practical questions that make this probability even less:

1) Vote counting fraud due to hidden electronic fraud with no paper trail. If you really want to make your vote count, you should design electronic voting machines and make sure it counts A LOT.
2) Hanging chads on paper ballots
3) Recounts in *selected* districts chosen by political processes to bias the election slightly.
4) Other forms of after-the-fact biasing

Happy election day.

For photography it's important to get some consistency in color rendering on your display.

I just profiled my display with my brand new "X-Rite eye-one display LT" (colorimeter that sticks to your monitor) and the "Argyll" CMS software for linux/windows/MacOS.

I installed the profile and "OH YEAH" SOOO much less blue than before. Stupid default settings on laptop monitors. Why do they make them SO BLUE? On a desktop monitor you can set things, but on my laptop, it's just fixed on BLUE.

Idea for a new magazine:

"Subset magazine: academic living in the west"

Full of ads for Dover books, Apple computers, Mathematica and Matlab. Articles include pictures of professors in over-architected academic buildings with fancy interactive whiteboard technology, and spreads of gourmet food.

Article topics would include "Tesselating the square: a patio you can count on" and "An Idea House of the 20th century: a tour of Einstein's private cottage"

If you don't get it, you obviously have not read Sunset magazine.

Come and get them (safe for work and anywhere else, I promise).

I'm studying solid mechanics in tensor form in class... The class is very frustrating because it's very algebraic, and at the same time, not at all careful about the mathematics... so I have to both see a lot of tensor algebra that doesn't make much sense, and I'm not really sure what the right concepts are because the definitions and soforth are very handwavy and we don't really talk about concepts at all.

It also doesn't help that the whole class lecture series is done on a half sheet of paper under a camera, and then broadcast at low res over the internet....Basically you can't teach people anything by sitting in front of a camera and carrying out long complicated algebraic manipulations.

So I've been studying the concepts of vectors, tensors, and differential forms / differential geometry. I've got "Geometrical methods of mathematical physics" by Schutz, and "The Geometry of Physics" by Frankel. Neither one is ideal, since they both are really aimed at people interested in general relativity or similar topics.

Nevertheless, I'm trying to understand, for example, the strain tensor, within a differential geometry approach... actually preferrably within a differential forms approach... I'm not quite sure what the difference is really.

sigh. time for sleep I think.

Today the students in my Engineering Risk Analysis (Probability) class said that they wished that I could teach the whole class, not just the discussion, and that they understand the concepts a lot better after they hear my lectures. I was walking on air for about a half hour.

That's me.... err there. that's better (new version with better color balance).

This year I've been pretty successful in my business doing mathematical modeling, data analysis, and soforth. During the first half of the year I continued to work with JKA in walnut creek, analyzing information from some of the projects that they were working on before I left.

During that same time, Francesca and I were shopping for a house, and I developed a model for pricing during the strong downturn of the LA housing market. Now, the market has a huge bid ask spread, and I'm talking with our realtor who's also become a friend, about providing that modeling expertise to his clients so that they have some idea of how to choose a price to place bids.

Also, I've been working on a biological data analysis project for David Morton's lab at OHSU (where hazelett works). The model involves predicting the distribution of time-to-behavioral-change when various mutant maggots are placed in various hypoxic environments. It's a dynamic model using some basic ideas from differential equations, but it works exceedingly well so far!

Most recently, I've decided to add to my capabilities by investing in a Canon DSLR outfit. This will be useful for both macro photography and technical documentation alongside my biology consulting, and forensics, as well as artistic photography and digital imaging consulting. I think there is a lot of room for educating biologists about digital imaging which is so important to their research, as well as selling art photos locally in the Pasadena area.

It's funny, now that I've gone through a process of getting myself into Civil Engineering research, my business ideas from Endpoint Scientific Computing (a consulting concept I tried to do 5 years ago or so) have finally started to pick up :-)

here's how I finally got it so I could spéll oüt internàtional words on my linux laptop.

First off, I run XFCE4 as my desktop, so the keyboard GUI setting is pretty stripped down. The best way to set up your keyboards is to go to the xorg.conf file (/etc/X11/xorg.conf).

There I have the following settings:

Section "InputDevice"
Identifier "Generic Keyboard"
Driver "kbd"
Option "XkbRules" "xorg"
Option "XkbModel" "pc104"
Option "XkbLayout" "us,us"
Option "XkbVariant" ",intl"
Option "XkbOptions" "ctrl:swapcaps,grp:switch,grp:alt_shift_toggle"
EndSection

The important lines are XkbLayout where I choose two layouts, both are "us" and the XkbVariant line where I set the variant for each "us". The "us" variant "intl" has "deadkeys like ' and " and ` and ~ which combine with other keys to produce accented characters.

the XkbOptions line tells the system to swap capslock and ctrl (so ctrl is now just to the left of "a" on my keyboard) and then also to switch between groups when I press left alt + left shift.

to get a ç I type Right-Alt + ","

It lets me put all the major european diacritical marks on my text and yet still have a "normal keyboard" most of the time (Just press Left-alt + left shift to toggle between normal and intl)

If you don't like this exact layout, you can go to /usr/share/X11/xkb/symbols (on Debian systems with recent xkb-data packages) and you can modify the layouts yourself.



Hope that helps someone out there.

A good talk by Stephen Pinker at the TED conference in 2007 compares the ways of life and specifically the amount of violent conflict over the last several thousand years.

The only issue I have with this talk is whether violence is also decreasing in absolute terms? As population grows from say 1 Million to 10 Billion, if violence declines by a factor of 100, we still have 100 times as many people suffering violence...

Nevertheless, it's important to not be disillusioned about rates of violence in the past which were, he argues, much much higher.