The Hanover Collaboration

Though reasonably computer literate when it comes to using software and programs in daily life, I’ll admit I’m completely ignorant of what is going on behind those programs and inside my laptop. Which is probably the reason I’m so upset at what I went through two weeks ago.

I’ll save you the long version, but basically, after a lot of haranging Applecare informed me that my logic board had failed and needed to be replaced. After $330, it now has.

Though I love macs (I have a 2002 15” Ti Powerbook G4), I have to admit this is the second apple that has had a mother/logic board failure on me in the past few months (the other was the DOC Freshmen Trips computer a week prior to the arrival of the first `09’s.)

I understand when hard drives, disc drives, screens, keyboards, and operating systems fail: all have dynamic and therefore vulnerable mechanical or program-based components. Isn’t a motherboard just a large circuit board with a bunch of stuff plugged into it? So what gives? How does a motherboard fail to the point that it needs to be replaced?

Space Physics graduate student retreat. JS and BP are graduate students. RH is the director large, well funded research institution. CE reviews funding proposals for the federal government.

Discussion topic: funding your research.

JS: Well, I was just about to ask about exactly that. When I graduate next year, is there any reason I can’t apply for funding to do research, like, out of my basement or where ever?
CE: Well, you get a lot for that money that goes to your institution. They take care of your heat and power. They give you a place to do your research.
RH: Also, there’s a certain amount of credibility that you gain by having an intuition’s name after yours.
JS: Obviously that’s the issue. “Jon Shea” isn’t prestigious enough. But research is supposed to be about merit and discovery, and I think it would be wrong to bias the funding too heavily by the prestige of someone’s institution.
CE: That’s true. And we do give some, maybe 11%, of our money to non-academic institutions. While it is possible to provide funding to unaffiliated individuals, logistically and in practice it is much easier if they are incorporated.
BP: Say a researcher had already, by whatever means, achieved a given result in his research. Suppose he then takes that result, and puts it aside, in escrow. What if he then submitted a propsal for funding of the research he had already done, and then used that money to do some new research. At the end, he could set aside the new result, and publish the old one. Would you say that this behavior ethical, and would you say that this behavior common?
RH: It isn’t ethical.
JS: But what if, for whatever reason, someone makes a great discovery without any funding, certainly that person deserves at least the same compensation as someone who has no guarantee of making any discovery at all.
BP: And, you’d be using that money to make new discoveries.
RH: The US government doesn’t pay for discoveries. It doesn’t buy them from people. It founds people to do research that they wouldn’t otherwise be able to do.
JS: But the outcome is the same in the steady-state case.
RH: That doesn’t matter. If you make a discovery, and you want to sell it to the private sector, then fine. But that isn’t what the government pays you for.

The results are in. If you would still like to take the test email me the results or post them in the comments, and I’ll add them to this chart. This post will be perma-linked to the front page. The results are in no particular order.

I’m not sure if any of you heard the news today, but here in building 18 today the mood is very festive. The day is especially exciting for the grad students in my group (although not nearly as much as it is for those in Shrock’s) because one of the other winners, Robert Grubbs of the other institute of technology, is our academic grandfather! Yes, academic grandfather is an actual term; I didn’t invent it just to make my self feel special.

The chemistry these guys worked on is metathesis which utilizes transition metal catalysts to form/break double bonds. I used the technique recently to make a polymer, and its ease of use and effectiveness left me weak in the knees. If all chemistry were that good we would be building nano-flying cars instead of sticks, balls, and the occasional lever or wheel.

Congratulations, gentlemen!

p.s. There’s a Dartmouth Connection here as well. Prof. Grubbs of Dartmouth is the new Nobeloriate’s son!

Solar flare affects communications, disruptions possible

This is pretty much exactly my research, and now likely to be the topic of discussion at this weekends CISM graduate student retreat.

My lab group’s work is on the MIT homepage today!! This research isn’t the sort of stuff I work on (I work on actuating polymers), but its damn cool. Congrats to my labmates and my boss!

Most engineering profs and most books with which I have come in contact are pretty lousy with notation. I am writing this entirely due to frustration with just such a book and just such a prof.

Physicists on the other hand, tend to be much cleaner. Why do you think? You might guess it has to do with the physicists have a much cleaner job. That isn’t it. If you see the same topic covered in a physics and engs textbook in the same depth, the notation in the engineering book is almost always less clear.

A decent theory is that the history of physics is so broad that differing conventions abound. That requires that each writer be clear about which convention he is using. Is phi the azimuth today or is theta?

It seems to me that in engineering there has been a lot more unification. This leads writers and lecturers to eroniously believe that I will recognize their notation or that I will remember what N_0 stood for from the lecture last friday.

Some of you who took certain classes with me might suggest a different theory. Engineers don’t understand that there is a concept behind that symbol. It is a variable to which a number is assigned when I plug it into the formula. While this was true of the students in those classes, it is not true in general of the people writing these books or teaching those classes.

Computer Science has a great concept here that makes this really simple: scope. Your local notation (as opposed to global like basic math operators) must be declared and goes out of scope when you start doing something else. Scope certainly doesn’t bridge chapters. It shouldn’t bridge from chapters into problems. If you go a few pages without using it, just refresh it.

Suppose that you want to refer to the speed of light, c. Whoops, I did it right. I mean that you want to refer to c. I already gave it away. If you just say c you make me guess from the context what we’re talking about. Sure sometimes it is easy to guess, but I should have to.

You should look at the crap I trying to read. I’m going to the bibliography to find the books from which the author poached this notation.

... but upon computing |A|² = A₃², the average radiated power per unit solid angle is (see Problem 8-7):

Eq. (8.111) power per angle = [complicated equation offered with out derivation or explanation]

Problem 8-7: Derive Eq. (8.111).

I was sitting in on a lecture by the esteemed professor “Phan”: who was talking about non-linear function approximation. In particular he was covering Nueral Nets and their applications. He made a point which, while silly, I found very exciting. I’ll try to explain why.

Nueral Nets

First let me quickly describe the math version of the nueron. A math nueron takes a linear function of its inputs and then passes that through a non-linear function. The non-linear function is predefined, e.g. tanh.

The exciting thing about a (math) nueron is that it is to some extent modeled after a brain neuron. The brain neuron recieves inputs from other neurons on dendrites. These inputs are “gated” by electrical and chemical regulators in a way that could conceivably be represented as a scaling. Then something happens in the middle of the neuron. The first bit, the linear function, summing the weighted inputs is probably not terribly accurate. But then we talk about neurons “firing.” I can do “firing” that with predefined non-linear basis functions no problem. So that part could conceivably be well modeled as well. Then it sends a signal down the axon to some other neurons.

I say that it is at least a workable model for the brain, considering that it has been proven that 1) a (math) neural net can be arbitrarily descriptive 2) by putting loops into the net, we can have memory.

Then also think about the fact that if we were to simulate the brain, it would take some 10^14 (yes thats one hundred trillion) parameters to describe it. Thats a large number but I might get to see computers with memory that big before I die; maybe even computers that could do computations with such a network.

Learning Chaos

The interesting thing about a neural network is that it can be “trained” to approximate any non-linear function. “Training” merely consists of a mathematically systematic way of tuning all of the parameters so there is an optimal match (or perhaps just locally).

Now we can train the network to model something unpredictably non-linear or chaotic. Then the network would be able to reproduce the signal, but only over the input range. If we were to try to extrapolate beyond the input, there would be unpredictable results.

At this point, Phan (jokingly) brought up the question of creativity. What if I trained the network to reproduce the first half of a song by Bach and then asked it how the rest of the song went? I would certainly get something. Under the right circumstances, I might even get something that sounds like music. I might even get something that sounds like Bach! Of course, I am the first person to dismiss the phenomenon of “emergent behavior” (it is just another way of calling a “bug” a “feature”). But if I were to design a program that could compose “Variations on Bach,” that would certainly pull some more people into Turing’s computability camp.

Then I remembered how Brayton would often start whistling a bar of some classic rock balad, and then go off into a noodle-land that still remained vaguely coherent. Maybe that’s chaotic extrapolation. I doubt it though. He does the same thing in Karaoki.

I’ve been putting off my EM waves homework for far too long now. My petty insistence on having “physical understanding” of the math I’m doing is starting to really get in the way. This happens a lot in higher level physics, but this time it isn’t just because the math is getting hard.

Electricity and Magnetism is broken.

Busted. Obviously, and irreconcilably unphysical.

At first, it looks like classical mechanics did. Yeah, there are some problems, but you suspect that quantum mechanics will fix them. Except, for E&M, it doesn’t fix them. As far as I can tell, every 20 years or so the Giants bat around some pet theories, and argue about some numerical coefficients. And then they go back to working on parts that aren’t broken. No sense in drawing attention to the failures.

Your first E&M homeworks have a lot of simplifications in them. So do your second homeworks. In the middle of your third class, you start to get a feeling for how deep E&M goes. And at the end of your third class, they ask you to break it.

I’ve never before been so upset with physics. I feel betrayed. Before, I thought I had something fundimental about the universe in my hands. Darcartes could go to hell, because I had the d’Alembertian. But it turns out this isn’t truth. It’s bullshit.

The bullshit happens to be right to about 1 part in 10²⁰, though. And that’s pretty good. (As long as you don’t look too closely at the wrong thing. Like how charged particles accelerate.)

Joe brought up the body count number game again. I get sincerely angry when people pooh pooh the civilian deaths.

What Lancet actually found was that 98,000 Iraqis had died since Operation Iraqi Freedom began. In a country of 25 million, that isn’t hard to believe. The 98,000 includes asthma, cancer, accidents, car crashes, heart attacks, old age, suicide… The 98,000 includes everything.

The average Iraqi lives to be how old? We’ll leave that as an exercize.

Even if we use the 21,000 number (the sum of the deaths reported by news agencies) that alone makes violent death due exclusively to the {insurection, rebellion, terrorism, Americans} a worse risk than every other way of getting killed in Iraq. Even if this were instead of ordinary murders, which certainly aren’t decreasing, it would be an enormous increase over pre-war Iraq. And yes, it would also be an increase over the state of California.

I know that you can handle the math.

How many times has the following happened to you?

You are in a restaurant with a bunch of people and it is time to compute the tip or divide the bill and people are intimidated about what to do. Inevitably the operation falls on you, the “math whiz”. You’re like, “What makes you guys think that I am any better at (multiplication/division) than any of you? I know someone has a cell phone with a calculator.” (Now to be honest, I am quick enough at number operations that I usually just do it for them. But if I think that the group is too math-intimidated, I say that as a public service.) The thing that bothers me is that my phone has both a calculator and a tip calculator. Why would there be both? You know what I’m getting at?

I have worse stories but some characters might be recognizable so I won’t share them.

I was sitting in class today faced with a terrifying mixture of calc, stats and linear. I couldn’t help but think that I was somewhat under-prepared. My first thought was to blame my previous teachers and the whole education system in general.

As the applications evolve, the education must be supple to keep up. People have been claiming for 40 years that mathematics education is to static and too rote. I can’t agree more. Perhaps it sounds funny (as in my favorite Tom Lehrer song), but “the important thing is to understand what you’re doing, rather than getting the right answer.”

Ask yourself where you spent the most time learning and where you spend the most time in application. If you’re like me, you spent 14 years learning how to compute things. But now I seldom actually add or multiply or divide actual numbers. I just write it out with the understanding that it could be done. I don’t often integrate stuff, I just write out what the integrals are. I don’t actually solve systems of equations, I just know when they can be solved.

When was the last time you used long division? Or even added 2 digit numbers?

The pinnacle of the standard mathematics education these days seems to be the integral. Don’t think that I am trying to downplay its importance but it is essentially long division again. Conceptually it is of critical importance but computationally it is a big waste of time.

When I tutored kids in vector calc I used this problem a lot.

d/dt(Something) =—divergence(Something)

I would say, “You already understand this equation. You’ve known it all your life. Give me examples of Somethings that makes this equation true.”

If we are over that hurdle, I don’t really care if they compute it wrong for some field or other. Alas, typically they pay me to increase their grade (as opposed to their smartness) and we end up spending most of our time computing it for this field and that field.

The nice part is that conceptually, most of math (the practical stuff, at least) is pretty intuitive. I’m convinced emphasizing the intuitive aspect would make learning easier. That, in turn, would free up time to cover more math. Maybe there’d be time to add the exponent and logarithm to the operations that the average American knows about. Maybe cover stats a little better and a little earlier. Maybe cover discrete math at all before sophomore Computer Science. Maybe make it to the convolution operation a little earlier. Maybe I wouldn’t have been hit up with quite so much new material today.

Since a number of you guys are or have been (or very nearly were) teachers, I am very interested in what you think about this.

Remember that crazy general in “Dr. Strangelove” who was so opposed to flouridation of water. Well, in the interest of full disclosure, I’ve got to tell you that I am not psyched about it either. What to know why?

Simple: I consider the original research to be inadequate.
What little research was done was on NaF while what is being put in the water is H2SiF6, an industrial by-product.

In light of the fact that flourine is very toxic (particularly in that nasty acid form, Scott, back me up here) and is largely retained, I would want some higher quality science applied to it. I just want to be sure that it doesn’t do any of the bad things that a lot of “crack-pots” suggest that it might do. Some of the crack-pots are pretty credible.

So what is the big problem with putting a freeze on water floridation while we get some higher quality data?

I don’t really like it when people say, “Evolution is just a theory.”
Technically they are right but it is very solid theory. There is a mountain of evidence in favor of it and zero admissible evidence in favor of it’s alternatives.

So it is with a little hesitation that I have to point out that “Global warming is just a theory.”

There was a discussion on the radio about the scientific concensus on the issue. The theme the host was hitting up was along the lines of, “Global Warming is a sure, proven thing. It is just simple physics and all the scientists agree. The only reason that a person would be skeptical is because of disinformation from the industrial lobby.” The scientist were a bit better (they tend to distinguish themselves from talk show hosts that way) and restricted themselves to talking about the sort of evidence they have like tree rings, sediment patterns, glacier cores and paths and what the evidence suggests.

“There is a simple and compelling theory as well as a great deal of evidence. Why shouldn’t I be convinced?” you may ask.

Go ahead and be convinced but don’t overstate yourself. Here is the concept we are trying to get our hands on: variation in global average temperature as a function of atmospheric content. Neither of which are easy to measure. As we go back in time, the data rapidly gets much much thinner. This is still very similar to the fossil record. The difference is that if I find one transitional species, that constitutes very powerful evidence for evolution existing. If you show me a temperature curve for only a single location (e.g. a glacier) that constitutes relatively weak evidence for global, as in everywhere, warming.

Let me digress a second and talk about the big worry about positive feedback. There are a few potential positive feedback loops available to this sort of ecological “crisis.” On of the easiest to grasp is that snow reflects sunlight. Less snow means more solar absorbtion which leads to more heat and less snow.

Let’s suppose that this sort of mechanism could lead to some sort of unrecoverable catastrophe. Why hasn’t such a catastrophe already occurred? The very same glaciers tell of pretty substantial temperature fluctuations. I suppose that if there were an uncontrolled positive feedback mechanism, then it would have been triggered by now. Since that hasn’t happened, it is suggested that there is an overarching negative feedback that regulates this problem.

That said, we are putting nature to some tests that it probably hasn’t seen before.

Back to my mean thread. The evidence right now is very strong that both CO2 and temperatures have increased for at least the last fifty years, with the Mauna Loa graph being the most popular. We also know the mechanism by which CO2 can increase temperature. There is an abnormally high level of CO2 in the atmosphere. Since people are a major source of CO2, we should maybe try to do something about it.

All of the efforts that I keep hearing about are passive ones e.g. “don’t burn coal.” Sounds good but coal is really cheap way to get heat which we can use to cook food, heat buildings and make steam to turn turbines that generate electricity. Looking into alternatives seems wise to me.

But what I really want to hear about is some more active ideas. The glacial record suggests that the CO2 level has risen and dropped pretty dramatically in the past. Some active engineering might be able reproduce such a drop. Think about it.

Running the numbers, putting shiney stuff on our roofs might not quite do it. But feeding ocean plankton might be a smart way to go about it. There are a ton of substances that absorb the shit. I don’t think it is unreasonable to imagine an industrial-chemical solution a little more agressive than eco-cement.=

I haven’t really thought about this very hard yet, let alone researched it. Go ahead, set me straight.

Dartmouth prof, Dan Rockmore chatting on The front porch last tuesday and I didn’t have time to warn everyone.

It turns out that he is on the radio all the time.

It sounds like an early lecture from a freshman seminar that gives away QDS’s for real cheap.