Objekt Orient

60m east-europeans less

2009-07-07T20:08:00.005+02:00

Europe’s 2008 population of 736 million is projected
to decline to 685 million by 2050 because
of its low country-level TFRs and in spite of
continuing net immigration. The decline, however,
is expected to take place primarily in eastern
and southern Europe. Eastern Europe’s 2008
population of 295 million is projected to decrease
to 231 million by mid-century, while southern
Europe is projected to decrease from 155 million
to 150 million.

http://www.prb.org/pdf08/63.3highlights.pdf

Some thoughts on that:

Austrian economy, that highly gained from the new EU member states, will suffer this demographic shift.

Western and Northern Europe are stable, but require higher automatisation of semis-skilled labor.

More semi-skilled immigration from the south - racist parties gain but society is to old for Nazi-cruelties.

US still dominates the century, due its founding myth its open to skilled labor immigration and is highly competitive.

India, Brazil are the new force, if China doesn't make it to become a mature service economy before it grows too old. Russia will decline.

The Law of Small Numbers

2009-06-29T08:10:00.002+02:00

In this groundbreaking work Ladislaus von Bortkewitsch shows that rare events are Poisson-distributed.

http://www.archive.org/stream/dasgesetzderklei00bortrich#page/n5/mode/2up

The Science of Better

2009-06-25T22:13:00.003+02:00

Damn, it was Operations Research what we were looking for all the time.

http://www.scienceofbetter.org/

http://en.wikipedia.org/wiki/Operations_Research

About the power of innovation

2009-06-07T13:57:00.005+02:00

Advocates of innovation describe it as necessary tool in a recession to regain prosperity.

Shed no doubt on it, but this is a macro-economics truth. For a company it does not make sense at all: the innovation bears a risk and we know from the market equilibrium theory that the risk surcharge is such that the utility equals non-risk investments.

A good theory but do I know my risks? Donald Rumsfeld once made an epistemological speech about unknown unknowns. The probability of innovation taking off, in the end is unknown. But theory behave as they would know, they just don't know that they don't know. So in the end its a Rumsfeld-ian "unknown unknown"?

As always, listen to Mr. Taleb carefully. He designates revenue from innovation in the fourth quadrant.

Flippant Juror

2009-06-05T19:26:00.004+02:00

During lunch I exercised on the "Flippant Juror Problem"

a jury of 3 jurors decides by majority
two jurors are serious experts, they make a right decision with the probability p
one is a jerk, he just flips a coin with probability for head q=1/2 to decide

Is this jury better than a single person who also makes a right decision with the probability p?

The solution is they are equal. Thats indeed counterintuitive! The jerk sabotages the jury.

Annotation: If they would all flip a coin with majority vote it would be also equal to one flip

This leads to a question related to the electoral process with majority vote in general.

Given N voters, how many "coinflippers" M it takes to reduce their probability for right decision to one decision, hmm try to solve that....

Architecture is about the 'Form follows Function' principle

2009-06-01T12:18:00.004+02:00

Functionality should determine the Structure.
If hardware would be perfect, if CPUs would process instantly an infinite amount of input for an extremely complex algorithm. If connectivity would allow for instant access to an infinite amount of data. If our programs could be proven right, never crashed. If nobody could eavesdropping or tamper our system, if one programmer could change the system in every direction at no time... we would not need to think about an architecture. The internal structure of a system would be of no concern.

The limits of given technologies demand counter-measures to support functions, therefore we employ architectural patterns and tactics to circumvent undesired limitations.

Not all technology hurdles can be circumvent in this way, and some functionalities remain unimplementable. Thus a quantum computer would allow more functionality but demand new architectures.

In the next post I will outline the Quality Attribute approach, to design the architecture of a software system out of business functions, given technological constraints

Somebody becomes rich just by luck

2009-05-22T21:08:00.036+02:00

Whenever one is concerned with rare events, events with small probability of occurrence, the Poisson distribution shows up in a natural way. - Falk, Hülser, Reiss

Lets spread some rice corns on a table, each cell represents a person.

Do we expect equal wealth for everyone, given that all have the same virtue and skill?

The Poisson distribution tells us 16 persons get nothing, the probability that you are one of them is 26%.
Over fifty percent get zero or one corn - means they are poor.
Thirtyfour percent are "two or tree corn" people, forming a middle class .

A person on the sunny side of life get four and two very lucky ones get even 5 corns, They are the rich five percent of society

#corn	#people	P(X=corn)
0	16	26%
1	14	35%
2	10	23%
3	6	11%
4	1	4%
5	2	1%	upper-crust

Random distribution of wealth leads to substancial inequality. Some people just make it, while most struggle. Is this the secret formula of "the american dream"?

Representativeness Heuristic

2009-05-17T19:51:00.006+02:00

Finally I concede myself to start with "Judgment under uncertainty: heuristics and biases" even though not finished with the development part of the Bayesian Network book.

The introductory chapter starts promising: the first cognitive antipattern (bias) discussed

Representativeness

When people are asked to judge if:
A belongs to class B
A originates from B
A follows from B

they use representativeness, or similiarity ommiting factors that should influence our judgement

Steve is very shy and withdrawn, helpful, but with little interest in people. He has a need for order and passion for detail.

How do people asess if Steve is engaged in a particular occupation (for example farmer, salesman, librarian, physician)?
They compare the description with the stereotype of e.g. a librarian.
Now have a look at the first fallacy with that: we ommit prior probability of outcomes, in fact we should consider that there are many more farmers than librarians before using the stereotype approach.

We fail to take into account prior probability, especially when given worthless evidence, neither wrong nor suggestive but simply worthless. With prior knowledge of a population of two laywers per engineer and the description

Dick is a 30 years old man. He is married with no children. A man of high ability and high motivation, he promises to be quite succesfull in his field. He is well liked by his colleagues.

The experiment shows that overall, the sample judged the chance of Dick being an engineer is fifty fifty!
Prior probabilities play an important role in Bayesian Networks, thus fixing our biased judgement.

"Over Christmas, Allen Newell and I created a thinking machine"

2009-05-05T02:59:00.005+02:00

A quote honoring Pittsburgh and its "son" Herbert Simon, who also said (in 1971!): What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention. Actually I desperately need his biography, he seems to be a crucial proponent with the study of human decision-making, behavioral economics and AI.

Best lecture I saw today.

Vacation and Training

2009-03-26T08:22:00.003+01:00

I hit the road until mid may, coming back with more architecture related topics.

Unmasking the Beard

2009-03-15T12:11:00.009+01:00

I trained my Bayesian Network modeling skills with a sportsbetting fraud. We call it a beard, when a professional is masking his insider abuse by placing his bets with help of an acquaintance. It will be in nearly all cases a relative or a friend. How can we imply from evidence of surname to being relative? And then from being relative to being "beard?" Or: how can we imply friendship, and then beard?

The naive model presented shows reasonable behaviour.

First the presentation of the marginal probalilities (the numbers) and the inference sensitivity (red boxes) of nodes

Being a beard can only happen if there is a pro to mask. So this evidence remains set. Friends and relatives highly influence the other nodes because of their outgoing edges and their prior conditional probabilities. Interestingly being from same district has also a high impact because its relatively rare measured within a nation. We see, being from the same district increases beard probability +3% in this model:

Download and play it with GeNie.

A screencast:

Preview of Wolfram Alpha

2009-03-10T13:32:00.003+01:00

You can today test a similiar system to the pre announced Wolfram Alpha and get a feeling of the potential. Ask quantitative questions like: How many legs has a chair? How old is Michael Jackson? Now think that WA will actually try to compute stuff: how much bigger is the GDP of US to china? I'm excited!

Regression Towards the Mean

2009-03-09T20:49:00.002+01:00

The psychologist Daniel Kahneman referred to regression to the mean in his speech when he won the 2002 Bank of Sweden prize for economics.

“

I had the most satisfying Eureka experience of my career while attempting to teach flight instructors that praise is more effective than punishment for promoting skill-learning. When I had finished my enthusiastic speech, one of the most seasoned instructors in the audience raised his hand and made his own short speech, which began by conceding that positive reinforcement might be good for the birds, but went on to deny that it was optimal for flight cadets. He said, "On many occasions I have praised flight cadets for clean execution of some aerobatic maneuver, and in general when they try it again, they do worse. On the other hand, I have often screamed at cadets for bad execution, and in general they do better the next time. So please don't tell us that reinforcement works and punishment does not, because the opposite is the case." This was a joyous moment, in which I understood an important truth about the world: because we tend to reward others when they do well and punish them when they do badly, and because there is regression to the mean, it is part of the human condition that we are statistically punished for rewarding others and rewarded for punishing them. I immediately arranged a demonstration in which each participant tossed two coins at a target behind his back, without any feedback. We measured the distances from the target and could see that those who had done best the first time had mostly deteriorated on their second try, and vice versa. But I knew that this demonstration would not undo the effects of lifelong exposure to a perverse contingency.

http://en.wikipedia.org/wiki/Regression_towards_the_mean

Bailout Micropayment

2009-03-06T19:33:00.005+01:00

A thought doesn't leave me. Why does no central bank issue iCash?

A innovation in online payment would lease the liqudity/ loan problem and boost consumer spending. Why not subsidize every computer with a smartcard reader approved for financial transactions?

Don't mix Correlation with Causality

2009-03-06T17:23:00.007+01:00

The statistic course or to put it in another way: even if there is a correlation between storks and babys in burgenland, storks are not the cause of their birth.

Why we should base our decisions on probabilistic networks

2009-02-21T08:13:00.008+01:00

We know from behavioral economics that humans are inherently biased in their decisions.
Basically "the expectations we have in our intuitive system are different than in our reasoning system" (D. Kahneman).
Constructing a probability network enables us to use our reasoning system with the qualitative structure of the influences in our decisions, while leaving the actual inference to the conceptual framework of Bayesian statistics. Because this is were usually our intuitive system kicks in and biases us towards wrong decisions. Because we are build to survive in nature, not to reason about the risk of complex derivative finance products.
Having a probabilistic network documents our decisions and allows a Shewart-cycle of improvements.
We would be able to peer review our decision networks and building up a pattern language of optimal decisions. As Steven Wolfram tried for the natural sciences by collecting algorithmic particles describing nature, we could try to make a executable Wiki of fundamental decision patterns describing recurring decision problems.

Influence Diagrams

2009-02-15T11:00:00.013+01:00

A Bayesian network is a probabilistic network for reasoning under uncertainty, whereas an influence diagram is a probabilistic network for reasoning about decision making under uncertainty. - Kjaerulff, Uffe; Madsen, Anders

An Influence Diagram consists of observations, decisions, utility functions and a precedence ordering. It extends the Bayesian Network by a sequence of decisions in time, their utility and their costs.

For example the Influence Diagram above shows the utility of treating apple trees under the observation that they lose leaf. This can be because of drought or sickness. If the phenomenon persist until later when the harvesting time gets nearer, we might lose income.
If we want to maximize income, a cure is barly recommended by the Bayesian inference. It will cost us € 80 and give us a gross income of € 176. We earn € 96 over the € 84 gross for net without treat.
Download a version of the apple tree problem without barren variables leaf' and dry'.
Download GeNie

Understanding Bayesian Theorem

2009-02-07T14:01:00.008+01:00

Eliezer S. Yudkowsky says:

Maybe you see the theorem, and you understand the theorem, and you can use the theorem, but you can't understand why your friends and/or research colleagues seem to think it's the secret of the universe. Maybe your friends are all wearing Bayes' Theorem T-shirts, and you're feeling left out. [...] What matters is that Bayes is cool, and if you don't know Bayes, you aren't cool.

His first example in a nutshell (by M. H. Herman):

Here's a story problem about a situation that doctors often encounter: 1% of women at age forty who participate in routine screening have breast cancer. 80% of women with breast cancer will get positive mammographies. 9.6% of women without breast cancer will also get positive mammographies. A woman in this age group had a positive mammography in a routine screening. What is the probability that she actually has breast cancer?

Only around 15% of doctors get it right (Casscells, Schoenberger, and Grayboys 1978; Eddy 1982; Gigerenzer and Hoffrage 1995; and many other studies.)…On the story problem above, most doctors estimate the probability to be between 70% and 80%, which is wildly incorrect…The correct answer is 7.8%, obtained as follows: Out of 10,000 women, 100 have breast cancer; 80 of those 100 have positive mammographies. From the same 10,000 women, 9,900 will not have breast cancer and of those 9,900 women, 950 will also get positive mammographies. This makes the total number of women with positive mammographies 950+80 or 1,030. Of those 1,030 women with positive mammographies, 80 will have cancer. Expressed as a proportion, this is 80/1,030 or 0.07767 or 7.8%”

I modeled this as BN with GeNie and came to the right result:
yes = 7.8% genie displays it rounded to 8%
I learned hereby:

I don't understand Bayes well yet
I don't need to understand Bayes to model a Bayesian Network successfully

Advertisment as a game of chance

2009-01-30T19:07:00.008+01:00

Let there be no doubt: Ads are a gamble. Their purpose is to increase the chance someone buys your product. I realized this only when i red Seth Godin's thought: "If your ads work throw all the money on them." I think he failed to understand that there can be no determinism between ads and turnover, since an ad is never (1) forcing someone to buy or (2) informing that there is a product fulfilling a desperate need. In both cases: p(ads)=1

What we see is: there is a correlation between ads and turnover. Especially you don't know if the campaign works until you try it, a.k.a. subjective probability. Therefore we should model advertisement as a probabilistic game. Therefore the Kelly Criterion is highly relevant.

Thank you for listening

How should I call it?

2009-01-29T11:19:00.007+01:00

Did you know that german is the only language that preferrs notebook over laptop?

Monty Hall Problem

2009-01-25T12:44:00.012+01:00

The Monthy Hall Problem shows how little intuition humans bear to solve probabilistic enigmas.

Tree doors, one hides a price. The candidate chooses one and then the show master opens another one bearing no price. Would the candidate increase his probability in choosing the leftover door, giving up his first choose?

The astonishing answer: the two doors left (the first-chosen and the remaining-closed) are not of the same probability the price-doors. But the remaining-closed now has a probability of 2/3 in comparison with the first-chosen 1/3.

The trick: the show-master opened a door, given the candidates first choice. in 2/3 of the cases he had to avoid the price-hiding door, indicating that the other bears the price. only if the candidates original try was right, the show-master could open any other door.

I modeled this as BN with GeNie:

The candidate chose State0, the quiz-master showed State1 and the price probability resulted in p(Door0)= 1/3 and p(Door3)= 2/3.

Download the original Genie Ziege.xdsl.

Belief Nets

2009-01-18T17:21:00.007+01:00

Based on the Bayesian Theorem we are able to model a network of evidences and influences to conduce a prediction: a Bayesian Network or Belief Net (BN). Somebody found out, such BNs beat other techniques in predicting certain sportsbets.

BNs allow an expert to map his expertise as causal dependencies leading to a certain prediction.

While Humans have shown skills in doing so, there is evidence that they fail in computing the overall estimate well.

While the domain expert and data available to us build up the nodes and their node probability table (NPT), the reasoning is left to Bayesian Theorem. Machine Prediction with a human designer.

A second benefit is the documentation of the evidences and influences leading to prediction and decisions. The state of the art is in combining Decision Analysis and Belief Nets in the model.

With University of Pittsburgs GeNie modeling tool, the world of Belief Nets opens for experimenting also to the interested layman.

Lets share some experiments!

Belief

2009-01-10T11:39:00.006+01:00

The Bayesian Theorem is one of the most powerful black magic tools of mathematical occultism. I want to formulate in plain text the purpose and properties of Bayesian Theorem.

Given an evidence, to what degree can we assume a hypothesis is true?

Lets assume

Evidence e: a bettor places a for him unusual high amount on a horse

Hypothesis h: a bettor has inside information

Problem p(e|h): When we know e, to what degree can we assume h?

Hypothesis h is a little hard to prove. We would need a court to judge that there was a inside information or even manipulation of the game. Lets relax h to a more broader term and reformulate the problem. When a bettor places a unusual high amount what is the chance that he wins? We relax h to:

Hypothesis h1: a bettor wins
Where h is part of h1, i.e. a insider needs to win to be an insider.

We could look at what we know about our bettors:

Certainly we know when a bettor places an unusual high amount, lets say more that 5 times the average in a similar bet. We know how many bettors do that, p(e).
We know how many bettors are winning in our horse races, p(h1).
We know how many past times winners betted high, p(e|h1) means literally: given that a person wins what is the likelihood he betted high? We know that from the past.

Now Bayes says we can predict from this data if a person will win if he is staking high.

p(h1|e)= p(e|h1) p(h1)/p(e)

Lets say p(e|h1)>0.5, then we know winners are usually staking high. This gives us evidence that they are confident to win. Why could that be? Reflect on this :-)

p(h1) in a game like horse race should not exceed 0,5 since all the money is in the pot, i.e. there are no more winners than losers and winners are just a few. p(e) will be also under 50%, just because otherwise the amount would be "usual". Whats important here: the smaller p(e) in comparison to p(e|h1) p(h1), the greater p(h1|e).

The chance a highstaker is winning rises when fewer people are betting high amounts, while the ratio of winners that betted high stays the same. That is certainly true.

What is when p(h1|e)>0.5? Then its likely that high stakes lead to big wins.

In the case of p(h1|e)<0.5 I would certainly bet with the bettor without spending a doubt on his reputation.

The Guy said it all

2009-01-04T16:06:00.001+01:00

Forecast from the bottom up. Most entrepreneurs do a top-down forecast: There are 150 million cars in America. It sure seems reasonable that we can get a mere 1 percent of car owners to install our satellite radio systems. That’s 1.5 million systems in the first year. The bottom-up forecast goes like this: We can open up ten installation facilities in the first year. On an average day, each can install ten systems. So our first year sales will be 10 facilities x 10 systems x 240 days = 24,000 satellite radio systems. That’s a long way from the conservative 1.5 million systems in the top-down approach. Guess which number is more likely to happen.

http://blogs.openforum.com/2008/11/25/the-art-of-bootstrapping/

Ads and Kelly

2009-01-04T13:09:00.008+01:00

Seth Godin is asking a good question: "do ads work?", and if, why you dont buy all of them.

I want to make a game of it to look at everything through my Kelly Glasses.

What Ads do is improving the edge or advantage you have on a coin where head means buy, and tail means not. The odds are turnover minus costs, lets simplifiy and account only for the marketing costs.

Kelly knows that in a game of chance it's no good idea betting you entire bankroll. You could go bankrupt within one throw. The same is valid for everey amount of investment greater than a certain amount f>f'.

If you could guesstimate your edge right, then G=edge/odds denotes your optimal investment rate