posted by
ben on
10.11.01 at 14:14, **null**, *null*, math, *math*, rant, *politics*, *rant*, *rave*,
Leave a comment

http://www.washingtonpost.com/wp-dyn/content/article/2010/10/22/AR2010102205451.html

So, so wrong. We need more math education. Yes, set theory may not be terribly useful (unless you interact with tuples on a daily basis), but calculus is a prerequisite for understanding the world around us. If we want to understand how many of the automated systems around us work, linear algebra and more is neccessary. The math education that comes with a typical college degree is totally insufficient to understanding the world.

That means only a select few can understand how things work, which in turn means that an even smaller select group can improve on the working of things. This is a problem. Wider math education means the ability of society to build more complex machines. That means great productivity and that means greater wealth.

Some education has value. Some does not. Math has value.

No hidden pieces. Black pawn direction doesn't matter.

A friend of mine found an interesting anagram of "BANACH TARSKI".

.

.

.

It's "BANACH TARSKI BANACH TARSKI"...

Comment from: Devin [Visitor] · http://www.hwaethwugu.com/blog

Fucking mathematicians. They come up with formal systems and then complain when they don't make sense.

Also, that the notion of area wasn't formalized until 1901 (and wasn't proved consistent until 1979) really pisses me off.

Also, that the notion of area wasn't formalized until 1901 (and wasn't proved consistent until 1979) really pisses me off.

The unprincipled approach to machine learning: use algorithms you've never even heard of in weka on your 100mb of stock data...

I'm having another one of those "Unless I've fucked up, I'm going to retire to Kopipi" moments. Let it stay.

Wow... that one only lasted a minute before I realized what I did wrong.

Anyone know how to invoke Matlab from the command line without bringing up the GUI? It seems to have something to do with the -automate flag, but I don't get it.

Right now I'm doing matlab -r computeBeta;exit

This brings up the Matlab window and closes it when done, but is kind of dumb. No one knows, do you? None of you use Matlab... boo.

Update: ...crap. Since it spawns off a new process, my Python keeps running happily while the Matlab runs and bad things happen. bah.

Weka seems to have gotten a hell of a lot cooler since I last used it (at least 3 years ago it seems). I suggest it for all you ML/data mining needs.

I don't know if this is a good idea... I currently have this phrased as a little game.

(1) When the day starts you get the open price and all the previous lows, highs, opens and closes.

(2) You then choose if you buy or do nothing.

(3) If you buy, you choose a sell price.

(4) If the stock hits the sell price during the day, it all sells and you make money.

(5) If it doesn't, the stock sells at the close price, which is usually catastrophically lower than the price you paid.

Does this seem reasonable? I'm not good at this game.

Histogram of highPrice(time)/openPrice(time) for AMD from 2002-2003. Note the spike at 1, or days on which the open price was the high price. On 26% of days, the high price is less than 1% greater than the open price. About 8% of the time, the high price is the open price.

On 52% of days, the high price is 2% greater than the open price. Does anything distinguish these days?

40% of the time, the high price never hits that 2% mark and the close price is less than the open. I don't like these days. I need to figure out how to avoid them. I suspect this is impossible.

It's a good solution, the only problem is that it depends on data from the future.

I don't understand how game theory and information theory fit together. All this momentum trading stuff has game theoretic aspects... for instance, should you sell with a 3% gain, or hold an extra few days for a 5% gain with some risk? I don't really know how to evaluate that, so I just do everything empirically... which is stupid.

Then, there are all these information theory questions, like given equity x went up 4% over the last few days, what is the probability of a 1% gain tomorrow... which seems like a notion that's tugging lightly as ideas like mutual information... since the question is really: is this Brownian motion, or something more systematic? Mutual information doesn't work as a metric, because the space the input series exists in is so huge that there's going to be lots of information, even though, for the intuitive notion I'm thinking of, there is in fact very little.

A good metric is hard to find.

So, I can evaluate those probabilities, find a bunch of joint probabilities, but how the hell does that fit in with these cutesy little strategies like holding a few extra days until the market wanders back to a profitable place?

That, and I don't know how to cross validate with time series data. I've been inching my way forward, giving it another year more data every once and a while to see what happens, which usually shows me that I've been over fitting like crazy. You can't test over different stocks, because they are incredibly dependent on one another.

Ack. I'm such a poser.

Beta for stocks traded on NASDAQ, NYSE and AMEX with market value over $1 billion and offerings before Jan 2002. Beta computed from 2002-2005 inclusive with VTI used to represent the market.

posted by
collin on
07.03.15 at 14:57, null, **math**, nonsense, *null*, science, *math*, *technology*, tech,
Leave a comment

Self assembling electrochemical sensors? [linky]

Anybody every hear of him? Sounds like Widrow's memsistors. It was Widrow, right?

Results

Searchstart: 3 The string 03081980 occurs at position 95,782,797 counting from the first digit after the decimal point. The 3. is not counted.The string and surrounding digits:

84481319032327216606 03081980 06677246175214953568

this query took 0.034847 seconds

And other random pi-ness.

posted by
collin on
07.02.26 at 09:01, null, math, nonsense, visual, **science**, *math*, *random*,
Leave a comment

Um, I don't see anything...

Am I right? Is there any practical use to this? Or is it just the same old; large enough data will have patterns of long period regardless of the structure of the data.

Abstract nonsense is a popular term used by mathematicians to describe certain kinds of arguments and concepts in category theory ...

Such and such is true by abstract nonsense.

I cried a little.

posted by
collin on
07.02.22 at 15:10, null, math, **nonsense**, *math*, rant, *random*, *rant*,
Leave a comment

I was looking at some random things...

Are the humanities guilty of making up nonsense words in a futile effort to have a semblance of logical, nee mathematical, consistency?

Bring it on bitches.

Dust bunnies are diffusion limited aggregates.

They were just lost in a bookstore. I'd like to think he would have laughed.

And some interesting heuristics.

Comment from: ben [Member] · http://ben.nonplatonic.com

“Good,” Wronoski recalled saying. “Now I donâ€™t have to kill myself.”

-Bookninja

It's official. I will stab them in the eye.

And here's some actual papers.

So tired.

Comment from: ben [Member] · http://ben.nonplatonic.com

What the hell?:

"We've just solved a problem that hasn't been solved for twelve hundred years - and it's that easy," proclaims Dr Anderson having demonstrated his solution on a whiteboard at Highdown School, in Emmer Green.

"We've just solved a problem that hasn't been solved for twelve hundred years - and it's that easy," proclaims Dr Anderson having demonstrated his solution on a whiteboard at Highdown School, in Emmer Green.

-Ronny Peikert