I was introduced to Türkpop darling Serdar Ortaç by an eccentric Turkish-Brooklyner-Turkish waiter at a restaurant adjoining a budget hotel in Istanbul.  I buy a lot of music when I travel, and I often try to get suggestions from locals I meet.  The waiter seemed like a good person to ask, since – judging from his habit of publicly announcing the latest developments in his sex life to the restaurant guests at breakfast each morning – I assumed we were on pretty familiar terms.  After faking my way through a few painful minutes of Yankees-Red Sox banter, I got directions to the nearest record shop and the names of his favorite Turkish artists.  One of his suggestions was the guy pictured below, posing with his… uh, I guess that’s a domesticated panther.

Serdar Ortaç with unreasonably large domestic cat. The next person to make a "pussy" joke about this album cover is getting a pointy stick in the eye.

I left the store with Ortaç’s latest album Kara Kedi (“black cat” in Turkish, whence the…) and an album of remixes from a couple years ago.  I find that remix albums are a solid bet when buying music by an artist I’ve never actually heard, because if I end up hating the artist, I can still usually find a palatable remix.

Strictly as a matter of personal taste, I’m not really a fan of Ortaç’s singing.  Türkpop gravitates strongly in the direction of nasally belted lyric ballads, which ain’t really my thing.  His instrumentals are well off the hook, though – wacky strings and funked-out noodling brass all over some intense thudding bass.  Accordingly, I decided to do some chopping.  I took two tracks from Kara Kedi, extracted just the bits where Ortaç isn’t singing, and shuffled them all back together.  The resulting track is after the jump, at the bottom of the post.

Among the parts of each song that I did decide to leave in were several unexpected Jamaican rap interludes.  Where did those come from??  About half an hour of digging through YouTube and Google results revealed the answer.  The Jamaican vocals apparently come from tracks by Sahara, a collaboration between Romanian singer and producer Costi Ionita and Bulgarian singer Андреа.  (I think that romanizes to “Andrea,” but don’t hold me to it.  Her legal name is Teodora Rumenova Andreeva.)  You may know Sahara: the project enjoyed some minor fame in the US last year due to a collaboration with Bob Sinclair and Shaggy.

The person actually performing the vocals, however, is the Jamaican-Romanian artist Lenox “Buppy” Brown.  (Check out his website.  I mean, really check it out, especially that rad gif of a waving Jamaican flag.)  One of the tracks I cut up, “Sanırım,” uses essentially the entirety of the Sahara track “Tyalee,” with Ortaç singing interludes.  Here’s “Tyalee.”

And below is “Sanırım,” for comparison.  As you can see, in a sense, they really are the same song.

The source of the vocals on the other track that I used, “Poşet,” is a bit more mysterious, but they are still evidently by Buppy via Costi Ionita, as evidenced by an angry comment that Buppy left on a taken-down video of the song:

Oct 1, 2010 – I just want the public to know that This Is Lennox Buppy Brown, the JAMAICAN voice that is on this track, SERDAR ORTAC & COSTI IONITA is not playing fair game. Why using my voice, lyrics and faking my image along with not putting my name to represent my voice? You guys are users and unfair artist representation. You both should be a shame of yourselves as musician & artist. Be real and true to the public.

I’m not one to make normative claims regarding IP ethics, so I will only confirm that Buppy is not mentioned anywhere in the liner notes for Kara Kedi, although Costi Ionita is seemingly credited, as “Ionita Constantin,” on both “Sanırım” and “Poşet.”

Finally, here’s my edit:

None of these involved any drawing or what would traditionally be considered technical artistry.  It’s all just math.

Ultimately, the easiest one to make turned out to be ItsYourDime’s edible peanut butter play dough recipe on Instructables.  Soon, however, I realized that this recipe actually presented a challenge: I love peanut butter, and if that’s what my putty was going to be made of, I wanted to be able to eat it at the end.  That meant that I was going to have to make it conductive without making it gross or poisonous.  I needed a conductive but edible ingredient to blend into the peanut butter.  The solution, of course, was graphite powder.  Graphite is flavorless and it’s just carbon, so eating it shouldn’t have any harmful side effects; after checking the materials safety data sheet to make sure, I set to work.  The recipe is after the jump.

Recipe:

Ingredients: 2 parts honey, 3 parts peanut butter, 4 parts dry powdered milk, 5 parts graphite powder.

Directions: Mix ‘em up.  That’s it.  You’re done.

Be careful not to inhale the graphite or get it in your eyes.  The end product looks kind of gross, but it actually tastes pretty good.  Don’t eat too much of it, though: graphite isn’t harmful to eat, but it’s not meant to be food.  If you want to store the stuff, put it in tupperware or wrap it in something air tight and put it in the fridge.

I’m still working on improving the recipe.  As it is, the putty still has a lot more resistance than I’d like, and I’m looking for other blends or ingredients that will make it more conductive.  I’d also like to make it look a little more appetizing.  Drop a comment if you have any ideas.

Okay, so it’s not really a full-on polygraph, since all it measures is galvanic skin response, and in any case polygraphs can’t actually detect lies.  But at least it looks kinda cool, if I may say so myself.

The LED on the front shows the wearer’s GSR.  It shines in varying shades of purple: the more blue it is, the higher the wearer’s skin conductivity; the more red it is, the lower the skin conductivity.  The idea behind using GSR in lie detectors is that when the subject gets nervous they (supposedly) sweat slightly more, and so their skin conductivity increases.  As you can see, in the picture the light is shining bright blue, which means I am telling a filthy fib.

The shirt measures GSR through two electrodes in the shoulders.  The “electrodes” are actually quarters duct taped to some telephone wire I found in the basement.  I try hard to stay true to the ghetto-fab ethos.  Aside from that, the only electronics involved are the LED, some resistors soldered into a PCB, a bunch more telephone wire, and of course an Arduino.  The shirt itself I found on a discount rack somewhere in Kenwood mall and then spray painted; the window on the front is some plastic that I cut out of a folder from Staples.

So, in the end, the VeraciT won’t actually prove to your friends that you’re telling the truth.  But it will prove that you’ve got a well funky shirt.

Many thanks to the inimitable Josh Brown for loads of help on the project.

The button on the breadboard selects between the photoflexophone’s three modes: light, flex, and off.  The off mode is the most important, since the thing makes infuriating noises if you leave it on.  The other two modes allow you to produce different pitches either by shining varying amounts of light on the photocell or by bending the flex sensor.  If you change the flex sensor out for a thermistor, you can also play it by altering the ambient temperature in the room.  If there’s ever been an instrument controlled by anything less practical than ambient temperature, I’d desperately love to hear about it.

At the moment it plays notes in the A minor harmonic scale, but it can play in any mode or just a continuous range of pitches.  The sound all comes from a piezo, though, so it’s pretty terrible.

Ingredients: Arduino, breadboard, flex sensor, light-sensitive resistor, push button, piezo element, a few ordinary resistors, and a whole bunch of jumper cables.

The party celebrating the first grant by the London chapter was held upstairs at The Griffin pub near Old Street.  It was a lot of fun, but it was kind of suspenseful.  The five finalists for the £1000 grant had to give pitches to the ten AF London trustees, after which the trustees went downstairs to the bar to deliberate before coming back up and announcing the winner.  Here are the finalists, along with brief, highly inadequate descriptions of their projects, in the order they presented:

Mary: DIY fashion design for the tech-savvy end wearer, with an eye toward visualizing data in clothing.  Complete with a Roomba mod for laser-cutting cloth.

Nathan & Frederick: Web service and iPod Touch app to help independent journalists in Afghanistan publish and organize field notes and visual information.

Matt: Streetview Street Art, a site that documents street art in its spatial and temporal context using Google Streetview.  The site is still incomplete, and Matt needs funding (or the help of generous hackers!) to get it finished.

Oscar: “The Big Dipper,” a project in which Oscar puts artificial stars in the sky over London, so that people can see constellations in a city ordinarily too light-polluted for stars to be visible at all.

Adam: A sculpture of the Earth showing where and when people are being born and dying, with red and blue LED’s flashing at different rates according to birth and death rates in different locations.

All of the finalists’ proposals were truly awesome, so I apologize that my descriptions didn’t do them justice.  In the end, Oscar’s project won the grant.  Hopefully we’ll see new and improved artificial constellations over Hyde Park and Shoreditch by the end of the summer.  I really hope that the other projects end up working out as well.  In particular, Matt needs some people with web development skills to help get his site fully functional; if you’re a web hacker with some free time and you want to get involved with Streetview Street Art, let me know and I’ll put you in touch.

I’m really excited about Mary’s project too, but also deeply afraid.  Of course, everyone joked about her laser-Roombas being the beginning of the end, but only because we were genuinely scared.  For weeks I’ll be waking in a cold sweat, unable to shake the fear that I’m about to get cut into a DIY dress.

Just kidding, Mary.  I really hope you get funding.

The application deadline for the next AF London grant is June 30th, and the award party will be not too long afterward.  Apply for the grant if you have an idea, and come to the party if you’re in town!  I’m looking forward to it already.

Computational Methods for Bounding Chromatic Numbers of Graphs

1 Introduction
Many central problems in graph theory involve the process of graph coloring. A coloring of a graph is an assignment of a label, or “color,” to each vertex, such that no two connected vertices have the same color. Perhaps the most famous example is the problem of map coloring: a map determines a graph by assigning a node to each country, with an edge between two nodes whenever the corresponding countries share a border. A coloring of the graph then corresponds to a coloring of the map in which neighboring countries never share a color. Appel and Haken famously proved that for maps, there is always a coloring with no more than four colors [1].

In fact, for any graph G, there is a minimum number of colors needed to color G. This number is denoted χ(G) and is called the chromatic number of G. There are known bounds on the chromatic numbers of certain classes of graphs. For instance, Appel and Haken’s result is the statement that if G represents a map (a so-called planar graph), then χ(G) ≤ 4. However, for a general arbitrary graph, the only known way to calculate χ(G) exactly is to find an optimal coloring, and the only known algorithm for optimal coloring is to enumerate every possible coloring of the graph and find a best one. That process is immensely computationally intensive; for an n-vertex graph, enumerating all colorings requires up to O(n^(n+2)) steps! It is therefore highly impractical, in many cases, to find the exact value of χ(G), and so we turn to approximations.

In this paper, we will explore a few computational methods for placing upper and lower bounds on χ(G) for random graphs without attempting to calculate it exactly. If we can find a coloring of G that uses n colors (an n-coloring) then we know that χ(G) ≤ n. To this end, we will explore two methods for coloring graphs in ways that do not necessarily use the smallest possible number of colors: the greedy coloring algorithm, which runs in at most O(n^2) steps; and the method of independent sets, which gives better results but can require up to O(n2^n) steps. We can also put a lower bound on χ(G) by finding cliques, or complete subgraphs, of G. χ(G) must be at least as large as the size of the largest clique in G, since no two vertices in a clique can share a color. Finding the size of the largest clique in G takes at most O(2^n) steps.

We will test the effectiveness of these algorithms on samples from two families of graphs. We will primarily use graphs from G (n, p), the space of random undirected graphs of order n, where the probability that there will be an edge between any pair of nodes is p. The second family of graphs that we will use is Gk (n, p), the space of random undirected graphs of order n such that there is never an edge between vertices vi and vj if i ≡ j (mod k), and the probability that there is an edge between any other pair of vertices is p.

In the following section, we will discuss the greedy coloring algorithm. We will then address the problem of finding cliques in a graph, after which we will use the same clique-finding algorithm to derive the independent-sets coloring algorithm. Finally, we will derive the complexities of the algorithms used in this paper.

The trend was found specifically in Indian, Chinese, and Korean families.  Families from those countries might prefer to have boys for a number of reasons: they have patrilineal cultures, in which the male children carry on the family name; raising male children is often seen as a good investment, since they are more likely to find gainful employment; and especially in India, raising female children, for whom parents may eventually have to pay a dowry, can be a financial burden.

There are several ways for couples to ensure the birth of a son.  In Asian countries, sex “selection” is most often accomplished by aborting female fetuses, or through female infanticide.  In the United States, families are more likely to use pre-implantation methods, such as sperm sorting followed by artificial insemination or in-vitro fertilization, although sex-selective abortions are still performed.  Whatever the method used, sex selection is problematic.  For one thing, the desire to weed out daughters is misogynistic, and suggests that these parents have some unfortunate ideas about gender roles.  For another, the practice of artificially selecting a child’s sex amounts to treating the child as a vessel for the parents’ expectations, rather than as a person in his or her own right.

Artificial sex selection needs to stop, but the question of how to stop it is a difficult one.  My position as an advocate for total reproductive freedom is that it is unethical to regulate many of the technologies that are used to predetermine the sexes of infants.  In particular, a woman’s right to get an abortion must never be subject to any conditions.  That means that we can’t legislate an end to sex selection.  If we want to solve the problem, we have to address its cause: the underlying beliefs that give rise to the preference for sons.

That’s a much loftier and tougher goal than outlawing sperm sorting, and it’s not clear how to go about accomplishing it.  Certainly the US would benefit from more widespread education on gender issues, for everybody.  But would that be enough to change a set of beliefs that many people take as cultural norms?  What else can we do to get people — not just Asian-American immigrants, but everyone — to see their daughters and their sons as equals?

If you have any ideas, leave a comment.

The beloved Shallow Goodale Shower

But the other day, all of that changed.  I realized that I was being stupid.  My fears had no basis in reality.  Unless the drain is clogged, there’s no reason to expect that a shower will ever overflow, for the simple reason that the rate at which water flows through the drain is proportional to the water level.  As more water fills the shower, its own weight pushes it out of the drain faster and faster.  If the drain is clear and reasonably large, the water should stop rising long before it overflows, because it will be flowing out at the same rate that it is coming in from the showerhead.  Since it’s IAP and I had nothing better to do, I thought I’d work out the math and measure the rate of outflow through my shower drain as a function of the water level.

Theory:

Suppose V is the total volume of water in the shower, and A is the area of the shower floor.  Then the height of the water is V/A, and the rate of outflow through the drain will be kV/A, where k is some constant of proportionality with units of area divided by time.  If r is the rate of inflow through the showerhead, then we have:

dV/dt = r – kV/A.

This is an extremely elementary equation, and we can see by inspection that the general solution should be:

V = C * exp(-kt/A) + Ar/k,

where C is some constant that depends on our initial conditions.  We don’t actually care what C is, because it is part of the transient solution.  That bit with the declining exponential will die out – that’s why the shower doesn’t actually run over – and the water volume will eventually stay fixed at Ar/k.  That’s the steady-state solution.

Experiment:

Our goal is to determine k!  In order to do that, we need to measure r and the steady-state value of V/A (we don’t need to know both V and A as long as we know how they’re related).  To measure r, the output of the showerhead, I needed to find some container of known volume.  I rummaged around for a while and the only thing I could find was an empty can that had once held cononut milk.  It would have to do.

The cononut milk can-cum-scientific instrument.

The can held 754 ml, and the shower filled it in about 2 seconds, for a flow rate of 0.377 l/s.  Then I let the shower run until it was about as full as I thought it would get, and stuck a ruler into the water, which came up to about the 1.5 cm mark.  That’s all the data we need!  (1.5 cm) * k = 0.377 l/s gives k = 0.251333… cm^2 / s, which when we account for sigfigs and wild inaccuracies is more like 0.3 cm^2 / s.

Well.  That was fun.

Lately around campus there have been a lot of signs encouraging me to “Close your hood sash when not in use!”  I do not know what a hood sash is, exactly – I believe it is some sort of equipment that chemists use – but I do know that the signs are telling me to close the sash when the sash is not in use, rather than when I am not in use.  That’s interesting.  Consider the following sentences:

1. “Close the hood sash when not in use.”
2. “The robot closes the hood sash when not in use.”
3. “Birds collide with the plane when flying.”
4. “Collide with the plane when flying.”

In #1, “when not in use” clearly refers to the hood sash.  Syntactically, my judgment is that it could also refer to the subject, but that meaning would sound very strange.  In #2, however, the phrase could refer to the hood sash or the robot, with a slight preference for the robot.  For #3 I have only a slight preference for “when flying” modifying “birds” rather than “the plane,” but I have a strong preference for interpreting #4 as “Collide with the plane when you are flying.”

What’s going on here?  In many cases, the meanings of the words themselves may force a particular interpretation.  Perhaps #1 can’t be interpreted as “when you are not in use” because it’s unclear what it would mean for the subject of the sentence, who is presumably a person, to be “not in use.”  In #2, the subject is explicitly identified as a robot, and since it’s easier to imagine what it means for a robot to be “in use,” both interpretations become viable.

Looking at only 1-3, one could conceive of a simpler explanation: “when not in use” can refer to any noun phrase that appears explicitly in the sentence; since the subject in #1 is implicit, the interpretation “when you are not in use” is not prefered.  That reasoning falls apart when applied to #4, though.  My theory is that any interpretation that doesn’t conflict with the meanings of the words in the sentence is viable, but that it is always somewhat preferable to interpret a phrase like “when not in use” as refering to the subject of the sentence.

This probably is nothing deep, but it has me puzzled.