Complexity Digest 2000.07

14-Feb-2000

For individual e-mail subscriptions go to Subscriptions.
Previous issue 2000.06 | Next issue 2000.08

Content

  1. A Tax Of The Worst Kind, Worldlink
  2. Market Force, Ecology, And Evolution, arXiv
  3. Why Onions Have More DNA Than You Do, The Harvard University Gazette
  4. Computation With Biomolecules, PNAS
  5. Revealing Uncertainties In Computer Models, Science
  6. Verbal Learning Following Sleep Deprivation, Nature
  7. Synchronization Properties of Brain Wa, ves, PNAS
  8. Contest: Detecting Sleep Breathing Problems From Heart Signals?, Ary Goldberger
  9. Rapid Climate Change, PNAS Special Feature
    1. The Heat-Salt Conveyor Belt Of Ocean Circulations, J. Marotzke
    2. The Tropical Pacific: The Sleeping Dragon Wakes, R.T. Pierrehumbert
    3. Ozone Depletion And Global Warming Synergy, D.L. Hartmann et al.

  1. A Tax Of The Worst Kind, Worldlink Next Article Bookmark and Share

    Corruption is a topic that has found little attention in the academic community and among complexity researchers. Its impact on a global scale, however, is horrendous. The close intermingling between economy, politics, and organized crime leads to unspeakable human suffering and destruction of the environment. When the large-scale atrocities in Kosovo were ended and the refugees could return some had the hope that this was a last horror-story from a war-torn 20th century. But the stories of human cruelties and suffering that we hear from Chechnya do not sound any more humane. We know that both in Yugoslavia as well as in Russia and other states of the former Soviet Union corruption was and is a number one problem.

    Corruption is a social phenomenon that thrives especially well in dark areas outside the public attention where many hidden agents interact based on emergent and changing rules, continuously adapting to new laws as well as technologies like the Internet. It seems to be an ideal topic for analysis from complexity theory. Rob Jenkins sees corruption as a "tax" on the public and on companies that undermines the business environment as a whole. (See also a Business Week Online Interview with Peter Eigen at the World Economic Forum, (ComDig 2000.4.1.2)).

    " Popular theories of the 1970s, which defended corruption as efficiency-enhancing, or at the very least a necessary evil, have been debunked by a mountain of research demonstrating its corrosive impacts. These include lost productivity, increased poverty, skewed public expenditure patterns and a host of other downstream ills."

    There was a hope that by introducing World Bank and IMF sponsored market oriented reform programs one could eliminate bureaucratic sources for corruption. It turned out, however, that corruption as a self-organized structure could easily adapt to the new situation: " Paulo Mauro, an IMF economist, argues that "the shift from command economies to free-market economies has created massive opportunities for the appropriation of rents [that is, excessive profits] and has often been accompanied by a change from a well organized system of corruption to a more chaotic and deleterious one.""

    In an information industry the early (and selective) availability of insider information can be extremely valuable and thereby a natural incentive for corruption: " The opportunities for corruption thus continue to evolve and in some cases far exceed the possibilities during the era of state control. Even where reform decisions are taken impartially, advance notice can be extremely valuable to private sector players, especially when regulations governing capital markets are concerned. Continuous reform makes inside information, on both timing and substance, a much sought after commodity.

    That the fate of both foreign and local firms hinges on these increasingly complex policy decisions highlights the importance of international factors in explaining why corruption has not only not vanished, but by some estimates actually increased. "

    Other critical factors are "Common border" situations like in Europe where crime organizations can focus on the "weakest link", a country with lax anti corruption laws or enforcement. Once drugs, weapons, illegal immigrants etc have been smuggled into this country, they can be moved with minimal risk to the other countries.

    On the other hand modern technologies also make it now possible for governments to respond and introduce new laws and way to enforce them. Most promising seem to be international agreements that make the flow of money more transparent and reduce the possibilities for money laundering.


  2. Market Force, Ecology, And Evolution, arXiv Next Article Bookmark and Share

    This paper evolved over a period of more than six years and summarizes some of the insights of one of the founders of the prediction company (see also the book reports ComDig 2000.0.13, 2000.5.12). It discusses the problems with equilibrium pricing models and introduces a non-equilibrium model of price-formation. It is a discrete time dynamical system ("map model") with additive noise. The general model can -depending on parameters- show fixed (equilibrium) points but also periodic or chaotic oscillations or it can "blow up" and grow indefinitely. Which solutions will be generated depends on the chosen trading strategy translated into model parameters. The interaction of market agents and their environments is seen in analogy to biological ecosystems with the price as emerging order-parameter that is generated by the activity of the agents but also influences the decisions of individual traders. He groups trading strategies into two broad groups:

    Technical or chartist strategies that only depend on the price history (for instance a trend following strategy)

    Value or fundamental strategies that are based on assessment of value in relation to price.

    He shows that value strategies lead to regularities in prices in the form of negative auto-correlations ("if the price was high yesterday it probably is lower today") whereas trend following strategies lead to the opposite sort of regularities namely positive auto-correlations ("if the price was high yesterday it probably is even higher today"). A combination of both strategies can make the price series have low autocorrelations, even though it has strong nonlinear structure.

    "A combination of value investors and trend followers gives rise to commonly observed market phenomena such as fat tails in the distribution of log-returns, correlated volume and volatility, and temporal oscillations in the difference between prices and values."

    The eco-system of these agents and their strategies evolve according to the capital that the agents accumulate and where they allocate it in reinvestments. The time-scales at which this evolution takes place is typically much longer than the typical time-scales at which individual agents operate. The resulting timescales at which markets progress towards efficiency can be estimated from the model and empirical data to be of the order of years to decades. Technical trading can then be seen as a way to exploit market inefficiencies that continue to appear in a non-equilibrium situation because of the generation of new information for instance based on innovations. In the biological analogy this would correspond to mutation and recombination.

    Farmer closes his discussion with the observation: "In the spirit of Dawkin's (1976) concept of memes, and E.O. Wilson's (1998) program of consilience, this paper develops an analogy between financial and biological ecologies. This analogy depends in part on the assumption that financial strategies are evolving automata. Strategies can be arbitrarily complex, but the key is that (like a genome) they can be regarded as algorithms that evolve with experience. This is in contrast to the prevailing view in neo-classical economics that the essence of a financial agent is the ability to reason and solve problems de novo. Of course, the behavior of real financial agents involves a mixture of both experience and fresh reasoning. My view is strongly influenced by my own experience as a practitioner: The investment strategy used by Prediction Company is based completely on evolving automata. Decisions are made entirely by computers without human intervention, using programs developed through an historical trial and error process. Of course many investors do use de novo reasoning, but my guess is that in the vast majority of cases, experience and culture dominate over freshly generated logic. There is a nonetheless a spectrum of possibilities between these two poles: By properly articulating both extremes, and exploring the middle, perhaps we can arrive at a view that correctly characterizes real people in real markets."


  3. Why Onions Have More DNA Than You Do, The Harvard University Gazette Next Article Bookmark and Share

    Everybody knows that genetic information in the DNA contains the instructions according to which cells from bacteria to humans develop and grow. It seems that nobody really knows why the DNA strands contain "junk", apparently meaningless sequences that seem to have no influence on the development of the cells that contain them. What is even more mysterious is that different species carry with them (and inherit it to the next generation) different amounts of junk. As a consequence the overall size of the DNA molecules for different species has an enormous variation: The largest DNA is about a hundred thousand times bigger than the smallest (eukaryotic) DNA). This is comparable to a situation where one edits, for instance, a web page in one text editor and it occupies, say, one kilobyte (kB) of disk space. Then if one edits the same page in a different editor, it would suddenly occupy 100MB. From this analogy one can see that some species add more junk to their DNA than the worst currently available html editors add to web pages.

    With such a variance in signal-to-noise ratio one can expect that the (useful) information contained in a DNA molecule is basically independent of its size. For instance the DNA that was in charge for assembling Einstein's brain was more than hundred times smaller than some of the DNA that build amoebas with a very limited repertoire of behavior.

    Petrov et al. studied one mechanism by which this paradoxical situation is maintained. They looked at the DNA of the fruit fly (Drosophila) which is relatively small compared to other insects and compared it with the DNA of Hawaiian crickets (Laupala), which have abut eleven times larger DNA strands. The mechanism they studied is related to spontaneous loss of nonessential DNA during insertion and deletion (indel) mutations. They could confirm the hypothesis that this spontaneous loss is much higher in Drosophila than in Hawaiian crickets and therefore. That indicates that this property is inherited and not randomly distributed. But the authors have to admit that they are still far from understanding why they are doing this and if there is an evolutionary advantage related to the specific size of a DNA in a specific environment.


  4. Computation With Biomolecules, PNAS Next Article Bookmark and Share

    Chemist, biologists, and even farmers can claim that they have been using "DNA computers" for a long time: In a sense any breeding effort to improve the performance of a plant or an animal can be interpreted as an approximate solution to a very complex optimization problem.

    Although the authors have not come up with a feasible general purpose DNA or RNA computer it is interesting how they could expand the range of interesting problems where one could imagine that parallel bio-molecular computation could offer new opportunities because of the astronomical number of parallel processors that can be implemented, at least in theory.

    Faulhammer et al. were able to specify a problem in RNA terms that corresponds to a variant of the famous mathematical problem of how many knights can be placed on a chessboard so that no knight is attacking any other knight on the board. One must admit that this problem does smell a bit like something that could be solved efficiently by chemical reactions. Also the general "programming" method that the researchers used looks quite familiar to what we know from ancient agricultural practices: Keep the individuals that performed well, eat the others.


  5. Revealing Uncertainties In Computer Models, Science Next Article Bookmark and Share

    Working with computer models can be very seductive: It generally is very easy to get numerical simulation results and answers to questions. In many cases the answer agrees with our intuition, it makes sense. If the simulated models become more complex it still is sometimes possible to perform a reality check by asking the "Does-it-make-sense?"- question. But in many situation intuition cannot be used as guidance and it can be tempting to trust the numbers or -even more seductive- fancy graphics or animation.

    The question of the trust-worthiness of computer simulations is as old as computer simulations. Sometimes a sensitivity analysis based on the variation of parameters and inputs and an analysis of the resulting output is sufficient. But if there are non-linear features in the model (as they always are in complex models) then a small combined variation of parameters can lead to results qualitatively different from those obtained by estimating error bars from single parameter variations.

    Cipra suggests calculating error bars and probability density functions for the results of complex simulations. A very severe problem for that approach is a curse of dimensions: The number of parameter combinations that would need to be tested grows exponentially with the number of dimensions and parameters of the model. It is clear that even with moderately complex models one can always (and easily) saturate the fastest available computers. Back in 1984 we did this with a model that simulated the impact of the strategic defense initiative ("Star Wars") on the strategic arms race. Ten million variations of a reference parameter configuration was all that a Cray supercomputer could handle in those days in a reasonable time (i.e. you want to be able to see the results when you come back from lunch). But those were sufficient to uncover sensitive conditions under which the model would predict any outcome in the range between zero and very large numbers.

    The use of genetic algorithms to search explicitly for parameter configurations that will lead to deviating results is another promising method for better than just statistical sensitivity analysis.

    As Cipra pointed out the situation is especially problematic in social science simulations where all numerical values are very fuzzy and hard to estimate with any level of precision. Nevertheless one can professional modelers that specify their parameters (and simulation variables) to a small fraction of the most optimistic empirical uncertainty and limit sensitivity analysis to a handful of scenario simulations.

    One can hope that the workshops mentioned in Cipra's article indicate the beginning of a more serious effort in developing quality control for the virtual world of simulations.


  6. Verbal Learning Following Sleep Deprivation, Nature Next Article Bookmark and Share

    It is known that sleep deprivation has serious effects on human mental performance. For instance after a night without sleep it is more difficult to learn new words.

    Drummond et al. were interested if there are any observable changes in the brain that would give an indication for the physiological source for this temporal impairment. They tested verbal learning performance of a number of volunteer subjects in a rested control state and after a night without sleep (35 hours total sleep deprivation). They could confirm a significant decrease in free recall performance but no significant change in recognition memory. Subjective sleepiness increased and the level of concentration went down, as one would expect.

    The authors used functional magnetic resonance imaging (fMRI) to measure the blood oxygen level in different areas of the brain and found that contrary to their expectation the activity of the prefrontal cortex (the location of working memory) was enhanced and not reduced after sleep deprivation. Other brain areas did show a reduction of activity after sleep deprivation and the authors interpret their findings as dynamic, compensatory changes in cerebral activation. In other words, it looks like other, more rested brain areas try to take over cognitive functions of areas that were affected by sleep deprivation. If this phenomenon also can explain hallucinations that are induced by sleep deprivation remains to be seen.


  7. Synchronization Properties of Brain Wa, ves, PNAS Next Article Bookmark and Share

    Electrical brain rhythms have traditionally been divided up into different frequency bands that are associated with different mental states. Best known among them are alpha rhythms in the 8-12 Hz range that are associated with a relaxed resting state (eyes closed). Beta rhythms (12-30Hz) are associated with alert mental activity whereas gamma rhythms (30-70Hz) have recently found a lot of attention in connection with the "binding process" during e.g. the perception of an object where different features are integrated into a coherent structure. There have been a number of theoretical models to explain the synchronization (and de-synchronization) of the tens of thousands of neurons that have to act together in order to be observable as an electrical EEG signal.

    Kopell et all. study a simplified neuronal model to demonstrate how different mechanisms can give rise to different rhythms. The model consists of two excitatory neurons (the activity of one neuron will facilitate the firing of the other neuron) and two inhibitory neurons (the activity of one neuron will inhibit the firing of the other neuron). The neurons are mutually connected with the connection between the excitatory neurons being turned off to produce gamma rhythms. To induce transitions to beta rhythms the authors introduce an extra "after-hyperpolarization" (AHP) current to the excitatory cells.

    The authors state that their model simulations are consistent with empirical observations that gamma rhythms are generally more spatially localized than beta rhythms because of the expected transmission delays between spatially more separate neurons. They concede, however, that gamma synchronization even between the two brain hemispheres (across the corpus callosum) have been observed indicating the high adaptability of neuronal conduction velocities with ranges from 1mm/ms to over 20mm/ms. One aspect that we did not find discussed in this paper is the phenomenon that the gamma synchronizations often are only very short-lived (in human EEG ~100ms; for a discussion and references see e.g. [Mayer-Kress, 2000]) a time-scale that is consistent with those during which for instance associations take place. It would be interesting to what model parameters naturally describe the length of a synchronization event.


  8. Contest: Detecting Sleep Breathing Problems From Heart Signals?, Ary Goldberger Next Article Bookmark and Share

    Worldwide Time Series / Signal Processing Competition

    Detecting and quantifying sleep apnea based on the ECG: A challenge from PhysioNet and Computers in Cardiology 2000

    Sleep apnea (intermittent cessation of breathing) is a common problem with major health implications, ranging from excessive daytime drowsiness to congestive heart failure and increased mortality. A number of studies have hinted at the possibility of detecting sleep apnea using only the ECG. Such approaches are non-intrusive, inexpensive, and may be well-suited for screening, but quantitative assessments of their accuracy vs. conventional techniques are needed.

    PhysioNet (a public service of the NIH NCRR-sponsored Research Resource for Complex Physiologic Signals) and Computers in Cardiology (an annual conference focusing on computer applications in clinical cardiology and cardiovascular research) challenge you to demonstrate the efficacy of ECG-based methods for apnea detection using a large, well-characterized, and representative set of data. This competition aims to stimulate effort and advance the state of the art in this clinically significant problem, and to foster friendly competition and wide-ranging collaborations. Participants will present their results during a special symposium at Computers in Cardiology 2000 in Cambridge, Massachusetts, 24-27 September, at which two winners will receive prizes of US$500.

    Visit http://www.physionet.org/cinc-challenge-2000.shtml for rules and data.

    • Submitted by Ary Goldberger

  9. Rapid Climate Change, PNAS Special Feature Next Article Bookmark and Share


    1. The Heat-Salt Conveyor Belt Of Ocean Circulations, J. Marotzke Next Article Bookmark and Share

      The basic mechanisms that drive the great Atlantic conveyor belt (Thermo-Haline Circulation (THC)): heavy (salty + cold) water sinks in the North Atlantic to great depths and flows towards the equator. Upwelling at low latitude and surface currents back to the North Atlantic close the circulation. It is not so widely known that this huge convective cell crosses the equator and extends at depths of more than four kilometers far into the Southern hemisphere where it meets the much smaller Southern hemispheric THC cell that transports much less warm water to high Southern latitudes.

      Today it is not completely understood what factors could trigger a termination of the THC although recent models do show hysteresis and multiple equilibria, which means that for the same condition the THC could either, be on or off depending on its history. This is a typical situation for complex, non-linear systems close to bifurcations.

      It appears that small density differences at the two poles can lead to a strong equator-crossing of the THC which will effect the global amount of deep sinking and therefore mixing of surface and deep water. This mixing is one of the main factors that determine how much atmospheric carbon dioxide the oceans can absorb.

      Among the indirect effects of a change in THC are changes in the global radiation balance (how much solar energy is absorbed and how much is reflected back into space). These changes can happen through changes in sea-ice, clouds, carbon dioxide, and water vapor in the atmosphere. Today, very little is known about the complex interactions among those coupled factors but it could be shown that chaotic dynamics plays a significant role even in current simplified models. Although this property puts strong limitations on long-term predictability of the modeled system it also provides some basis for hope that with the help of techniques from chaotic control theory one day it will be possible to mitigate or even take advantage of some of the climate instabilities. Considering the magnitude of the economic impact of even moderate changes in the global climate it seems to be evident that an improved monitoring and simulation network will be an excellent investment.


    2. The Tropical Pacific: The Sleeping Dragon Wakes, R.T. Pierrehumbert Next Article Bookmark and Share

      The past ten thousand years has seen two special phenomena: The Earth climate has been unusually stable and human civilizations emerged. Could it be that the latter was only possible because of the former? (There is evidence that the emergence of hominids in Africa a few million years ago was also triggered by climate effects with a resulting reduction of forested land area.)

      One important factor in stabilizing the climate fluctuations seems to be the "great oceanic conveyor belt" that kept pumping the equivalent of one Peta Watt (= 1015 Watt, the power of about a million nuclear power plants) of warm water to the North Atlantic. It could be show that every time a rapid climate change took place it was connected with a switching of the North Atlantic conveyor belt, technically known as "Thermo-Haline Circulation (THC)" (See also the article by J. Marotzke (9.1) and ComDig 1999.beta 2.10, 1999.beta6.2). This switching seems to be so sensitively dependent on detailed parameter configurations that current (coarse) general atmospheric circulation models could not reproduce this empirical feature. It appears to be one of the situations where the "butterfly effect" of chaos theory can play a role in the unpredictability of climate changes.

      The situation is very different in the Pacific ocean where large scale circulations are mainly driven by wind patterns and therefore do not affect deep layers of the ocean. But the formation of convecting air is also "tippy" in the sense that small changes in initial conditions can either switch the convection system on or off. One of the parameters that have an impact on the wind patterns in the Pacific is the retreat of the ice around the Antarctic. Pierrehumbert even points out scenarios under which one of the most robust wind patterns the tropical easterly trade winds of the Pacific -familiar to many seafaring generations- could be turned off. He concludes: " If one is tugging on the dragon's tail with little notion of how much agitation is required to wake him, one must be pre-pared for the unexpected."


    3. Ozone Depletion And Global Warming Synergy, D.L. Hartmann et al. Bookmark and Share

      The atmospheres of rotating planets shows large-scale self-organized spatial patterns that often show coherent oscillations. They are especially visible on Jupiter but they also play a fundamental role for the climate and large scale weather patterns on Earth. Some of the most prominent patterns are ring shapes around the polar axis of rotation (at constant latitude) that are formed by circum polar wind patterns or vortices. Their importance stems in part from the fact that they act as atmospheric barriers that extends from the surface to the stratosphere and that can trap for instance ozone depleting gases around the poles.

      While the ozone depletion and the resulting ozone hole are stratospheric effects surface climate is determined by weather phenomena that happen in the troposphere at lower altitudes. Both of them, however, can interact by changing wind patterns of the circum polar vortices.

      Thereby a synergistic effect that couples ozone depletion and global warming can occur that can have significant impacts on the climate of the coming century.


Also available in: Simple HTML format | TXT format | TXT format with links | Print