In this issue:
Peter Bryant
College of Business
University of Colorado at Denver
Denver, CO 80217-3364
pbryant@castle.cudenver.edu
303-556-5833
CSNA welcomes two new Board members, Olga-Cordero Bra~na of American University, and F. James Rohlf, of SUNY Stony Brook. Also, Stanley Sclove, from the University of Illinois at Chicago is the society's new Business Manager. We depend on the volunteer services of many people, and I'm grateful to them (and to the other candidates) for their willingness to serve.
By now many of you will have received the preliminary announcements and tentative schedules for CSNA '97 in Washington, D. C. The program is exciting, TWO short courses are planned, and the accompanying opportunities for professional growth and stimulation will be great. Mark your calendars now, and try to attend. I hope to see you there.
Stanley L. Sclove
Department of Information and Decision Science
College of Business Administration
University of Illinois at Chicago
601 S. Morgan Street
Chicago, IL 60607-7124
slsclove@uic.edu
www.uic.edu/~slsclove
Hello, everybody! Happy New Year!
I'm in the process of taking over from our preceding Business Manager, Dawn Iacobucci. Many, many thanks to Dawn for all her hard work for the Society, and for her patience in teaching me the new job.
On behalf of our President, Peter Bryant, I am pleased to inform you of the results of the CSNA election held at the end of 1996. Our two new Board members are Olga Cordero-Bra~na of American University, and F. James Rohlf of SUNY-Stony Brook. All CSNA terms of office are two years. Congratulations to Olga and Jim, and thanks to all who had agreed to run. Thanks also to the nominating committee, Martha Cooper (chair), David Banks, and Mike Windham. Finally, thanks to the Board members whose terms just ended, Doug Carroll and Larry Hubert. We look forward to continuing to work with all of you in various capacities in the future.
All the newsletters are posted and accessible at the CSNA homepage at the URL http://www.pitt.edu/~csna/. Check the page from time to time for updates and new issues. In addition, the CSNA bibliographic reference Service is also posted there. The Service, as well as the newsletters, are now therefore free and accessible anytime.
Membership fees are still $65 for regular members, $50 for retired members, $35 for student members, and $50 for affiliates (who receive the Journal of Classification but do not vote). Encourage your colleagues to join! We look forward to another fine year with CSNA!
F.R. McMorris
Department of Mathematics
University of Louisville
Louisville, KY 40292
frmcmo01@homer.louisville.edu
(502)852-6826
I want to call the attention of CSNA membership to a workshop that I am co-organizing at DIMACS (Rutgers University) in May on the topic "Exploring large data sets using classification, consensus, and pattern recognition techniques". This is announced at the end of the newsletter.
I know practically NOTHING about massive data sets but I feel strongly that classification theory will/should have important things to contribute to the analysis of such data. I hope that some of you will be able to attend and contribute ideas and perhaps talks. Next year is a special year on Massive Data Sets at DIMACS and there should be many workshops. Check the DIMACS web site: http://dimacs.rutgers.edu
CLUSTERS OF TALENT
David Banks, Dept. of Statistics
Carnegie Mellon University, Pittsburgh, PA 15123
banks@stat.cmu.edu
The most important question we can ask of historians is ``Why are some periods and places so astonishingly more productive than the rest?'' It is intellectually embarrassing that this is almost never posed squarely---I can think of only two articles (Gray, 1958 and 1961) and two books (Kroeber, 1944 and McClelland, 1961) that tackle this directly. But Gray is a mystic, Kroeber waffles vaguely, and McClelland veers off into a fascinating but incomplete assessment. The question has never been the focus of professional attention in social history, although its answer would have thrilling implications for education, politics, science and art.
Geniuses are not scattered uniformly through time and space. Some
cultures have many more than one would expect, even after making
sensible allowance for imperfect records, biased perspectives, and
such gross factors as famine, war, and the magnetic effect of
libraries and patronage. Obvious clusters of geniuses occur in
* Athens, from about 440 BCE to 380 BCE,
* Florence, from about 1440 to 1490,
* London, from about 1570 to 1640.
If the reader agrees that these three societies show such a
remarkable excess of creative accomplishment that explanation is
demanded, skip on to section 2. The remainder of this section is
just a borderline pretentious argument to convince rational
skeptics that random chance is an inadequate explanation for the
intellectual inhomogeneity that history records.
First, the three cities and times listed above are not unique--- there are many other concentrations, although these are particularly conspicuous. One can spend pleasant postprandial hours noting similar clusters in Weimar, Paris (twice), London (the Romantics), Vienna, Japan (late Heian period), Persia (just before Genghis Khan), the T'ang dynasty, and New York, at times that I hope most readers can discern for themselves. My sense is that there is a continuum of remarkability, from the three stellar cases listed first through the slightly humbler collections indicated in this paragraph, and the degree of remarkability shades imperceptibly into average societal behavior.(There are also major vacuums in intellectual achievement---the Dark Ages were notoriously weak, and one should recall Orson Welles' comment that 500 years of Swiss peace produced only the cuckoo clock.)
But let us focus on the first list. Were general citizens asked to name famous Athenians, the handful of names produced would come entirely from the indicated period. Even were an academic interrogated, the list would surely lean (list?) towards this period (Euripides, Sophocles, Aeschylus, Aristophanes, Plato, Socrates, Thucydides, Herodotus, Xenophon, Anaxagoras, Demosthenes, Pericles, Aspasia, Alcibiades, Praxiteles, Phidias, Protagoras, Aristippus, Isocrates, Lysias, Lycurgos, Polygnotos, ...). The suggested dates are obviously approximate, but I defy anybody to name an Athenian who amounted to anything that was born after 380 BCE (well, Kazantzakis will get a footnote, and if anyone were foolish enough to put forward Pseudo-Dionysius the Areopagite, I remind them that the extant writings are of Alexandrian origin).
So what happened to the cultural IQ of Athens over the last 23 centuries? The genetics didn't change appreciably (at least until 1398, when Nicopolis fell to the Turks). Why did such vitality stagnate?
Florence is almost as compelling. There is an early bump of productivity with Dante and Boccaccio and Giotto and Cimabue, but it faded out (perhaps as a subtle consequence of the the Black Plague). Then came a new lot, with Michelangelo, Leonardo Da Vinci, Machiavelli, Botticelli, Donatello, Politian, Mirandola, Lorenzo the Magnificent, and so forth. But before the birth of Cimabue (1240) and after the death of Galileo (1642), not much happened in the city. How was it that all its achievements were concentrated into such a relatively small proportion of its span? The question is particularly puzzling when one realizes that today, in all of Italy, the standard of education is higher, the promotion of merit is easier, and the population is enormously greater; nonetheless, there is no one that could seriously be compared to Dante or Leonardo or Michelangelo.
Elizabethan London is the third example. Marlowe and Shakespeare and Jonson and Raleigh and Bacon and Spenser laid the foundation for English writing, and there are a host of lesser luminaries whose hands helped. But their momentum ended with the coming of Cromwell (Milton stands alone). From the close perspective of college courses in the Restoration and Neoclassical periods, we can name many well-regarded writers (Dryden, Pope, Wycherly) before the Romantic poets burst forth, but the central point is that English talent isn't evenly sprinkled like pepper on potatoes; rather, it clots inhomogeneously.
If these litanies of names have not persuaded the skeptical reader that geniuses are sometimes superabundant, then a more formal debate is needed. To pursue that dialogue, I would ask such readers to decide how they would plan to calculate the numbers of geniuses they'd expect to see under their models, in order that we may compare their figures to the historical record. But for me, that smacks of straining at gnats when there are important camels to swallow.
There is much else one could say. It is interesting to compare the primary modes in which these societies operated (Athens did plays and philosophy, Florence did painting and sculpture, London did poetry and plays), and speculate upon the modern roles of television, performance art, and rap. Also, scientific thinkers do not seem to cluster to the same degree that arts and letters do, perhaps because science is necessarily accretive.
Previous Answers
Having posed the problem, let's proceed to a catalogue of some of
the stock answers that have been given previously. The following
five can be rejected out of hand.
* Hegel, in his The Philosophy of Right, proposed the
"Zeitgeist" theory of genius. But this, at best, is description
masquerading as explanation. As good empirics, we seek a more
specific understanding of the problem.
* Kroeber (1944) listed many cases of apparent clusters, and
lists all the obvious factors (economy, education, world
leadership, etc.) that might play a role. His perspective is
anthropological and psychological, and he weaves in a fashionable
amount of Toynbee. But he confesses that he is unsatisfied with
his treatment, and I concur.
* Gray (1958, 1961) believes that geniuses arrive according to
numinously perfect mathematical cycles. He identifies three
distinct periodicities, and argues that when all simultaneously
peak, high culture hits the jackpot; when all three trough, the
world's culture is comparable to a redneck high school's. I think
the man was a crank. (Readers may disagree with this casual
dismissal of his lifework, but I hate seeing mysticism written up
as science.)
* McClelland (1961) counted the numbers of achievement images
(e.g., ``the farmer wrests his livelihood from the soil'' vs.
``the farmer sows and reaps his sustenance'') that appear in
samples of text from periods of artistic creativity and periods of
artistic decline. He finds statistically significant differences,
argues that these reflect important psychological differences, and
speculates that the engine of cultural enrichment is emphasis upon
individual success.
* Asimov (1951) suggested that psychohistorical forces could
cause the cultural florescences, but, after all, this is only
science fiction. And he never specified the mechanisms that drove
Hari Selden's futuristic mathematics. Except for McClelland,
these treatments suffer from vagueness. We want explanations that
make sense, and which can be corroborated by specific historical
research.
What type of explanation is adequate? My guess is that high points in cultural history require the confluence of many factors; some of these are more important than others. When all or most of the factors coincide, then one has a Periclean Athens, Laurencian Florence or Elizabethan London. When only several factors combine, the cultural eruption is more humble---one gets Goethe's Weimar, or the Lake Poets. Things trail off gradually; if virtually none of the factors obtains, then we call it a Dark Age. From this perspective, the sought-for answer is a list of factors that facilitate/militate the occurrence of genius, with some understanding of their relative importance. In a crudely statistical way (retrodiction rather than prediction), one can test hypothetical factors by determining whether their presence is associated with higher measurements on some suitable index of a society's florescence.
In general, it is statistically (and epistemologically) impossible for the historical record to suggest the factors (that leads to overfit; epexegesis is deferred). The researcher must first make clever guesses, which can then be corroborated by the record. The next section describes factors that have been proposed, and various strategies for discovering plausible factors.
Searching for Factors
When I beard social historians at cocktail parties, they usually
dismiss the problem of explaining excess genius as complex and
ill-posed. But when coaxed into conversation, several ideas for
facilitating factors come forward:
* Prosperity. They submit that a florescent culture needs the
economic wherewithal to support the arts.
* Peace. They suggest that a climate of peace is also
conducive to philosophical, artistic and (perhaps) scientific
progress. (But recall Welles' comment on Switzerland.)
* Freedom. They believe that artistic freedom from state or
religious control enables new growth.
* Social Mobility. They think that when class distinctions are
relatively permeable, then there is greater incentive for artists
to excel.
* The Paradigm Thing. They suppose that when a new medium or
perspective arises, then art flourishes until the vein of
originality is worked out.
All of these are good ideas, and superficially plausible. But
most contradict the historical record.
To be specific, the prosperity suggestion fails for Athens, Florence and London. Athens spent its boom period in combat with Sparta; the income from the Delian League went to the fleet. Athenian farmers could not tend their crops (cf. The Acharnians), and such staples as grain had to be imported. Similarly, quatrocento Florence was poor compared to pre-plague Florence. The Medici bank had about half the capital of the Peruzzi bank in 1340, and Lopez (1970) documents other indications of reduced standards of living. A symptom of this desperation was the revolt of the populo minuto, which pushed the Medici into prominence. And Elizabethan London suffered ``dearness without scarcity'' (inflation); this fell most heavily on the aristocracy and the very poor. Then the wool trade collapsed, England entered ``the worst economic depression in history'' (Wilson, 1965), and Parliament anxiously debated means of averting a Bellum Rusticum.
Regarding the peace hypothesis, it clearly fails for Athens. Florence was torn by internal factions (e.g., il popolo grosso vs. il popolo minuto, the assassination of Giuliano de Medici, Savonarola). London had to contend with the Armada, the war with Spain in Holland, and internal religious dissent.
Regarding artistic freedom, the Athenian plays were written for religious festivals, and the prize was awarded according to the taste of respectable, pious, and civic-minded judges (this caused Aristophanes and Euripides no end of trouble). In Florence, art was commissioned largely by the Church, sometimes by a patron, and had to voice themes prescribed in the contract. In London, note that Shakespeare's plays avoid all mention of religion and contemporary politics; Marlowe and Jonson were similarly cautious (in literature, not in their personal lives).
Regarding social mobility, this hypothesis seems borne out by our three primary examples. Athens and Florence were both devaluing the aristocracy and promoting mercantilism. In London, the early part of the period clearly shows the rise of the middle class.
Regarding the emergence of a new paradigm, this is difficult to judge concisely. Much of the problem involves distinguishing a perturbation from an innovation. Did the introduction of a second on-stage character in Athenian plays represent a new paradigm? Was Plato's decision to write invented philosophical discussion a minor influence on the content of thought? Similarly, in Florence, painting and sculpture were well-established before the peak occurred, but the invention of perspective and the rediscovery of the classical period may have constituted a paradigm shift. Finally, in London, the key change seems to have been that small groups of strolling players discovered they could pack a hall in a city, and people would stroll to them. This enabled more elaborate props and larger companies, while pressing the need for a larger repertoire. But this kind of change is not especially Kuhnian in spirit, and the problem merits more lengthy consideration.
One could propose other factors. It seems to me that each of the three societies under consideration enjoyed a substantial military victory in the generation preceding their florescence. Athenians whose names shine today are reported to have prided themselves on being the sons of the men who fought at Marathon. Florence was not a military force (the Italian city-states relied upon mercenary condottieri in time of war) but in 1254 they conquered Pisa and Lucca. This secured an outlet to the sea, which was essential to their economic expansion. And in 1588, England conquered the Spanish Armada. This made the seas safe for colonial empire, and was a watershed for British morale.
Also, the great minds in each of these societies tended to hang out together. Socrates spoke with everyone; the playwrights talked shop, and the orators honed themselves upon each other. In Florence, artists trained under an apprentice system that pulled talents together, and Vasari describes frequent visits by the greats to each other's studios. Leonardo and Michelangelo held a public contest over The Battle of Anghiari; meanwhile, the poets and philosophers clubbed together at Lorenzo's mansion. In London, much of the theater circle met for drinks at the Mermaid Tavern, and one expects that their common profession ensured their lives crossed even more regularly. Aubrey reports that Bacon visited the Mermaid Tavern too, and doubtless Bacon knew Raleigh, who was sufficiently friendly with Marlowe to rise to the Shepherd/Nymph bait. Does the social intercourse of good minds produce great minds?
A third possible factor is education. In each of the three societies, education tended to be as personal as a punch in the nose. In Athens, the upper class had tutors and the lower classes shopped for their educations among various freelance teachers. In Florence, the upper class had tutors and the masses learned as apprentices. In England, the upper class had tutors and the commoners learnt to write plays and poetry from each other, insofar as I can tell. All three of these systems emphasize individual instruction over the currently popular cattle drive approach. And there is ancillary evidence (cf. the lives of Wiener, Maxwell, Dirac, Russell, Mill, Malthus, Arnold, Feynman, Tukey) that tutoring is enormously effective.
One can postulate many other factors. For example, it is suggestive that all three of Athens, Florence, and London had populations near 300,000. Also, all three had relatively democratic styles of government, and all three's florescences were ended by right-wing revolutions (the Rule of the 400, Savonarola, and Cromwell). Finally, each of the three were in the process of reinventing their language---Periclean Athens defined the conventions of Attic Greek, Dante made Tuscan the foundation of modern Italian, and the linguistic gap from Chaucer to Shakespeare is enormously larger than the gap from Shakespeare to us (but this could be due to selection bias, since language might gel around great writings, rather than great writings arise from volatile language).
There is never any shortage of hypotheses. The useful trick is to know how to test them. In this case, one could rank a sample of cities in terms of their cultural IQ, and then decide whether the hypothesized factor obtains for each of the cities. If the factor is more common for the florescent cities than for the average or below average cities, then the hypothesis is supported (this can be made formally statistical). To an extent, this style of reasoning is what is used in this section, except that I haven't elaborated the comparison by listing cities which have made meager cultural contributions.
The Individual
There is an alternative strategy for studying the problem of
clusters of geniuses. Instead of focusing upon the society that
produced them, one can study the minutiae of geniuses' lives,
looking for commonalities that might suggest cultural forces. For
example, if a study of many geniuses finds that disproportionate
numbers were tutored, then one might guess that creative societies
were those in which tutoring was a prevalent means of education.
Hayes and Simon (1985) report a study of composers. After sifting through much biographical material, they conclude that a minimum of ten years of serious study is required before anyone begins to produce important music. And subsequent research suggests that this ten year rule applies to many different disciplines, including mathematics, chess and poetry (though possibly not philosophy). Allen Newell is alleged to have proffered an explanation of this regularity:
"Greatness is relative, and humans compete against other humans. If intelligence is not a major factor, and if ten years represents the amount of time an exceptionally dedicated human is willing to invest, then, ceteris paribus, world class geniuses will be those who've marinated in a subject for ten years."
Clearly, the controversial element in this line of research is its underemphasis of the importance of native intelligence and skill.
In a similar mood, I undertook a study of 100 eminent men and women of Victorian science and letters. With the help of 11 undergraduates, I compiled a database that recorded 56 traits for each of the people chosen. We then applied a laundry list of semi-sophisticated statistical procedures to look for hidden patterns in the data. One particular question of interest was whether there were biographical traits that discriminated the artists from the scientists. To give a better flavor of the project, the first ten people in the database are: Matthew Arnold, Jane Austen, Charles Babbage, James Barrie, Jeremy Bentham, George Boole, Charlotte Bronte", Elizabeth Barrett Browning, Sir Richard Francis Burton, and Lord Byron.
Similarly, some of the biographical traits that were examined included birth order, family status, number of spouses, number of siblings, age at father's death, age at mother's death, whether or not the subject loved the father, ditto the mother, whether the subject was thrifty, whether the subject was gay, an estimate of the subject's sexual appetite, an estimate of the subject's precocity, a description of the kind of schooling the subject received, the intensity of the subject's religiousity, whether the subject had a sense of humor, whether the subject drank, the age at which the subject first produced good work, and so forth. Some of these traits were suggested by the research summarized in Sulloway (1996).
Unfortunately, the results of the analyses so far have been unilluminating. It appears that Romantic poets tended to have been raised by their mothers, and that scientists tended to come from happy, stable families. However, the statistical support for both of these conclusions is small (alpha approx 0.05), and the reliance upon undergraduates for the gathering of information ensures that, despite substantial efforts at data cleaning, the accuracy of the records is not beyond question.
Conclusions
The problem of excess genius is one of the most important
questions I can imagine, but very little progress has been made.
It surprises me that essentially no scholarly effort has been
directed towards it. I warmly solicit any suggestions from
readers that may help me to clarify my own confusion and
uncertainty regarding this.
References
Asimov, I. (1951). Foundation, Panther: London.
Gray, C. E. (1958). ``An Analysis of Graeco-Roman Development,''
American Anthropologist, 60, 15-27.
Gray, C. E. (1961). ``An Epicyclical Model for Western
Civilization,'' American Anthropologist, 63, 1036-1054.
Hayes, J. R. and Simon, H. A. (1985). ``Three Problems in Teaching
Problem Solving Skills,'' Thinking and Learning Skills, Vol. 2,
edited by J. W. Sega and R. Glases, Erlbaum: Hillsdale, NJ.
Kroeber, A. L. (1944). Configurations of Culture Growth,
University of California Press: Berkeley.
Lopez, R. S. (1970). The Three Ages of the Italian Renaissance,
University Press of Virginia: Charlottesville.
McClelland, D. (1961). The Achieving Society, D. Van Nostrand:
Princeton.
Sulloway, F. (1996). Born to Rebel, Pantheon Books, New York.
Wilson, C. (1965). England's Apprenticeship, 1603-1763), St.
Martin's Press: New York.
Contents:
Call for papers
Short courses announcement
CALL FOR PAPERS: 1997 ANNUAL MEETING OF THE CLASSIFICATION SOCIETY
OF NORTH AMERICA,
June 12-15,1997, to be held at the Mary Graydon Center/Butler
Pavilion American University, Washington, DC, USA
The 1997 annual meeting of the Classification Society of North America (CSNA) will be held in Washington, D.C., at American University on June 12-15, 1997. Two short courses, one on cluster analysis and the other on multivariate nonparametric regression and neural nets, are planned for Thursday, June 12. A welcoming reception will be held that evening. The regular meeting, including the CSNA business meeting, conference banquet, and regular paper sessions will be scheduled from Friday morning, June 13, until noon on Sunday, June 15.
CSNA meetings are traditionally informal and interdisciplinary, with few parallel sessions. Abstracts of papers that are presented will be distributed, but no formal proceedings will be produced. Conference attendees enjoy both applications and methodological presentations.
The plenary speakers will be Rob Tibshirani (University of Toronto) and John Hartigan (Yale University); Joseph Kruskal (Bell Laboratories) will speak at the banquet. There will be invited paper sessions on Environmental Applications, Inference on Phylogenetic Trees, DNA Fingerprinting, Challenges in Information Science, Classification in Public Health Medicine, New Problems in Biomedical Research, Clustering Problems in Marketing, and Problems in Classification. Additionally, there will be a session in which graduate students will present their research in the areas of classification and clustering.
The organizers of the meeting encourage the presentation of
contributed papers that cover a wide range of applications and
methodology that involve exploratory data analysis viewed in its
broadest sense. Papers related to all kinds of classification and
clustering problems are warmly solicited. Short abstracts of
papers should be sent to the Program Chair (David Banks,
Department of Statistics, Carnegie Mellon University, Pittsburgh,
PA 15213 USA, telephone (412)-268-2721, fax (412)-268-7828, e-mail
banks@stat.cmu.edu.), as well as any suggestions for symposia,
special sessions, topics, panel discussions, requests for further
information, or other contributions. The deadline for submission
of abstracts is March 31, 1997. The Conference Chair is Olga
Cordero Brana, Department of Mathematics and Statistics, American
University, Washington DC, 20016-8050, telephone (202)-885-3130,
fax (202)-885-3155, email olgacb@american.edu.
In connection with its annual meeting, the
CLASSIFICATION SOCIETY OF NORTH AMERICA
will repeat its well-received
SHORT COURSE:
INTRODUCTION TO CLASSIFICATION AND CLUSTERING
Thursday June 12, 1997
Mary Graydon Center/Butler Pavilion
American University
Washington DC, 20016-8050
This is an opportunity for students, faculty and other professionals whose work or research interests lead them into classification and clustering to get an excellent overview of the field, taught by some of the field's leading experts, and at a very reasonable price ($150 for industrial non-members, $75 for Government and Educational non-members, $50 for CSNA members, and $35 for students, including all course materials).
INSTRUCTORS:
Stephen C. Hirtle is Associate Professor and Chair, Department of
Information Science, University of Pittsburgh, Pennsylvania. He is
a member of the Board of CSNA and served as host and organizer of
its 1993 annual meeting. His research interests include
comparisons of clusters, similarity measures, lattice theory
models, spatial knowledge structures, and applications to
cognitive science.
Pierre Legendre is a Professor in the Department de Sciences Biologiques, Universite de Montreal. Montreal, Quebec, Canada. He is a member of the Board of Directors of CSNA and is one of its representatives to the International Federation of Classification Societies. He is well known for his work in cluster-analytic methods in ecology and biology generally, and is a Fellow of the Royal Society of Canada (Academy of Science).
Glenn W. Milligan is a Professor and Chair in the Department of Management Sciences at The Ohio State University, Columbus, Ohio. He has served as the Business Manager of CSNA from 1987 to 1990. His research interests in classification have focused on validating clustering procedures and on cluster recovery as affected by numerous factors. Projects have included studies of the "number of clusters" problem, clustering method selection, and variable standardization and weighting.
Preliminary Course Schedule:
8:30-9:00 (AM) Registration
9:00-10:30 Session I: Exploratory Data Analysis:
An Overview. Stephen C.
Hirtle
10:30 Break
10:45-12:45 Session II: Cluster Analysis.
Pierre Legendre
12:45-1:45 (PM) Lunch
1:45-3:45 Session III: Clustering Validation and
Comparison. Glenn W. Milligan
3:45 Break
4:00-5:00 Session IV: Example, Analysis, Discussion.
Milligan, Hirtle,
Legendre
Also, CSNA will offer a new short course on
Thursday June 12, 1997
Mary Graydon Center/Butler Pavilion
American University
Washington DC, 20016-8050
This class assumes a Master's level knowledge of regression (or substantial practical experience), and will be somewhat more technical than the other short course. It will cover, from a practical perspective, issues that arise in the use of new-wave computer-intensive nonparametric regression and classification methodologies such as MARS, Projection Pursuit Regression, Neural Nets, Loess, Additive Models; this entails discussion of related topics, including smoothing, the curse of dimensionality, concurvity, CART, ACE and AVAS, the bootstrap, and cross- validation. Most of the emphasis will be on regression, but classification will be treated as well. Attendees will receive copies of code with which to implement many of the procedures that are discussed. It is not necessary to bring a lap-top computer.
INSTRUCTORS:
David Banks is Associate Professor in the Department of Statistics
at Carnegie Mellon University, Pittsburgh, Pennsylvania. He is a
member of the Board of CSNA and an organizer of the 1997 meeting.
His research interests include high-dimensional nonparametric
regression, neural nets, graph-valued random variables, and
industrial statistics.
Chris Genovese is Assistant Professor in the Department of Statistics at Carnegie Mellon University, Pittsburgh, Pennsylvania. A recent graduate from Berkeley, his research interests include wavelets, inverse problems, functional magnetic resonance imaging (fMRI), and the practical application of computer-intensive methodologies.
Stephanie Land is an Assistant Professor in the Department of Statistics at Carnegie Mellon University, Pittsburgh, Pennsylvania. A recent graduate from Stanford, her research interests involve data fusion, new-wave nonparametric regression, classification, and applied statistics.
Preliminary Course Schedule:
8:30 - 9:00 (AM) Registration
9:00 - 9:50 Curse of Dimensionality; Smoothing
10:00 - 10:50 Generalized Additive Models;
Backfitting
11:00 - 11:50 Variable selection; Projection
Pursuit
12:00 - 1:00 Lunch
1:00 - 1:50 Neural Networks; Barron's Theorem
2:00 - 2:50 RPR, MARS, ACE, AVAS, Loess
3:00 - 3:50 Comparison of Methods
4:00 - 5:00 Guidelines for Applications
The fee schedule for both courses is the same: $175 for industrial non-members, $100 for Government and Educational non- members, $75 for CSNA members, and $35 for students. This fee includes all course materials.
For further information, contact
Olga I. Cordero Brana
Department of Mathematics and Statistics,
American University
Washington DC, 20016-8050
Telephone: 202 885 3130
Fax: 202 885 3155
e-mail: olgacb@american.edu
* MARCH 12-14, 1997: 21st Annual Conference of the GfKl, University of Potsdam, Germany. The conference program will include methods and applications of Classification, Data Analysis and Information Processing. Interdisciplinary aspects and interrelations between theory and practice will be particularly emphasized. For information contact Prof. Dr. I. Balderjahn, Lehrstuhl BWL - Marketing, Univ. Potsdam, August-Bebel-Str. 89, D- 14482 Potsdam, Germany. (Phone +49 331/977-3595. Fax +49 331/977- 3331. e-mail balderja@rz.uni-potsdam.de)
* MAY 4-9, 1997: ThRee-way methods in Chemistry and Psychology, Lake Chelan, Washington, USA. The meeting will focus on methods for analyzing 3-way and higher order data in the chemical sciences and psychology. The goal is to provide a relaxed atmosphere where ideas may be freely exchanged. Successes and failures will be discussed and opportunities for future research and application will be highlighted. For information contact Dr. Barry M. Wise, Eigenvector Research, Inc., 830 Wapato Lake Road, Manson, WA 98831. (Phone 509-687-2022. Fax 509-687-7033. e-mail 73633.2451@compuserve.com)
* MAY 29-30, 1997: DIMACS workshop on Exploring large data sets using classification, consensus, and pattern recognition techniques. Organizers: F.R. McMorris, Department of Mathematics, University of Louisville (frmcmo01@homer.louisville.edu) and Ilya Muchnik, DIMACS, Rutgers University (muchnik@lunar.rutgers.edu)
Workshop Theme and Goals:
Discrete mathematics (dm) and theoretical computer science (tcs) have become important tools in the areas of classification theory, consensus theory and pattern recognition (C2P). Looking at the types of problems attacked by the data mining community (as evidenced in the November 1996 issue of the Communications of the ACM), it is evident that cluster analysis (an area of classification theory) already plays a major role in massive data set (MDS) problems, though usually from a statistical point of view. It seems clear that dm/tcs-based C2P can play a role in this arena as well and that dm and tcs will eventually have important things to say in the analysis of very large, uncertain, heterogeneous data sets. This does not seem to be the case at present, so one of our main goals will be to explicate areas where dm/tcs-based C2P can play a central role (e.g., government, financial, ecological, astrophysical, molecular, etc.). This will be done in an exploratory workshop environment. There will be no specific applications that will be the focus of this workshop since we are looking for methodological approaches that involve, or might involve, the discrete mathematical aspects of C2P in MDS problems.
Participants:
Participants are expected to range from neophytes to experts in the field as well as practitioners who are users of C2P. All MDS talks will be expected to make some mention of C2P and/or discrete mathematics, broadly interpreted.
If you would like to give a talk, please contact one of the organizers. There will be a small amount of support available for participants who do not have other sources of support available. A list of speakers will be given as it develops.
Registration:
If you plan to attend, please send a registration form to Pat Pravato by email (pravato@dimacs.rutgers.edu). http://dimacs.rutgers.edu/workshops/classification/index.html
* JULY 21-24, 1998: 6th Conference of the International Federation of Classification Societies, Rome, Italy. Put it on your calendars - - - more information will be appearing later.