Pattern Recognition and Computing Concepts for the Biosphere
Allen Klinger
with contributions by Mark Burgin

Today we are faced with numerous ill-understood ecological, evolutionary or biological phenomena. Examples are global warming, amphibian disappearances, birds as descendants of theropods, and ethnobotany: medicines found by tribal civilizations. Pattern recognition, a field rooted in understanding how computers can act in situations where human decision-making is central, offers methods that can assist in dealing with such topics. For this abstract we summarize the range of phenomena to which we seek to apply our skills by using the term "Biosphere" as a shorthand symbol.

The fundamental issue in pattern recognition is set membership. In the Biosphere it is common to rely on a means for grouping that has dominated thought for more than two centuries, the system of Linnaeus. "Carl Linne', better known to the scientific community by his Latinized name Carolus Linnaeus, is often called the Father of Taxonomy. His system for naming, ranking, and classifying organisms is still in wide use today (with many changes). His ideas on classification have influenced generations of biologists during and after his own lifetime, even those opposed to the philosophical and theological roots of his work. ... specific details of Linnaeus's plant classification have largely been abandoned. Later systems of classification largely follow John Ray's practice of using morphological evidence from all parts of the organism in all stages of its development. What has survived of the Linnean system is its method of hierarchical classification and custom of binomial nomenclature." [1]

The nature of set membership was explored via an unusual method in work by the Russian mathematician Bongard. His one hundred problems [2] present a logical set exemplifying the complexity of human thought about grouping. In just six of the hundred two of us found these modes of organization: proximity, inclusion, size similarity/difference, quantity, number of inclusions, location of a general feature at top or bottom. All three collaborators are fully aware of the impact of Bongard's work. It has been one central consideration in the work of Douglas R. Hofstadter [3]. The issue that Bongard highlighted is of concept formation: in other words, grouping is associated with thought. For the collaboration we need to place this notion within pattern recognition and theory of classification.

Classification plays an important role in all fields of science. That process involves systematization of scientific information. It makes such information well-suited for human understanding, application, and transmission. Indeed accepted classification schema are the basis of specific scientific laws [4]. However, there is a vast wealth of human cultural experience from which pre-scientific thought distilled principles close to or identical with such laws.

There are two basic kinds of classification problems: direct, aimed at construction of an appropriate classification for a given set of data objects, i.e., a partition into sets or fuzzy sets; and inverse. In inverse classification the goal is to assign a given object to some class in a previously-established classification. Bongard extended Shannon's theory of information and communication to problems of pattern recognition. His fundamental achievement based information utility on more than probabilities, specifically using the nature of the problem under consideration. In applying this approach an essential component is an information measure: [5] developed both a general approach to information and means for measuring it; [6] proposes spatially-oriented measures of information. This work will involve inverse classification problems coincident with biological sensing pattern recognition.

In our work we will focus on three issues. The first is how to build on the two elements in the binomial nomenclature without accepting bias from past analyses. Here the idea is to accept computations based on a multiplicity of factors without first fitting them into the hierarchical Linnean system. The second involves a key aspect of pattern recognition that Bongard used in his pioneering work. This can be put most simply by observing that statistical methods are clearly quantitative but that many recognition decisions involve structure. That means that relation is as important as quantity. Hence we will investigate the following question. "What combination of relational and quantitative items best characterizes classification in the biosphere?" Our third activity involves pattern sensing in nature, both involving evolutionary/survival and cultural activities. We will work on environmental awareness of frogs in physiological and operational aspects of their ability to recognize audio patterns [7]. We will investigate multisite computing to deal with involving human cultural evolution [8] issues ranging over several disciplines. The focus will remain on survival (e.g., food production) enhancement through communications. We will consider human patterns and use computer systems to amplify specialist models to account for interdisciplinary views based on the overall model set membership classification and recognition theory.

Working Across Discipline Boundaries

The work is inherently interdisciplinary: it includes fundamentals in theoretical aspects of pattern recognition that involve both Klinger and Bergin in ongoing research that relates to human genome mapping. Specifically, the location of a 65-digit number satisfying certain binary constraints where until 1999 only one 10-digit numeral was known to do that. For detail on this issue please see [9].

Recent joint activity by Allen Klinger, Professor of Engineering and Applied Science, UCLA and Nikos Salingaros, Professor of Mathematics at UTSA (University of Texas at San Antonio) concerning information and order in spatial fields will be used to start an experimental process. This process involves a diversity measurement called life introduced by Salingaros and applied by him to many cultural areas of human endeavor. Please see [10].

Peter Narins, UCLA Professor of Physiological Science has an active program investigating the physiology of response to audio signal by biological entities. The electrical measurements of sound patterns is a fundamental topic in the UCLA Computer Science Department graduate course 276A Pattern Analysis and Machine Intelligence offered annually by Allen Klinger. Mark Burgin, Professor of Mathematics is capable of teaching and contributing to both this course and the audio signal research through his past activity. He and Allen Klinger began a recent effort to investigate the cultural and physical reasons for the importance of triad [11].

Evolutionary and survival relationships of species have been thoroughly described from a multidisciplinary viewpoint in [7]. We know that sensing and communication abilities were an essential step in human evolution. Further the ability to locate key information such as the availability of a food resource was an ultimate cause (term used in [7]) of the survival path leading to cultures becoming dominant. We seek to investigate signals' role in supporting domestication of plants and animals using pattern recognition and classification methodologies.

In [7] survival is described in terms of a culture making the best choice among different means of obtaining some overall success in garnering food resources. For example, low success rate from large animal hunting which yields a huge bounty on some days and nothing on others. This is compared with a comparitively conservative gathering strategy that yields on the average an aggregate amount of food great than that from hunting. Making a best choice between essentially random outcomes is similar to the pattern recognition and classification systems that use probability to place an unknown-category instance. Equally striking with these abstract issues are the range of human activities where pre-scientific societies accurately estimated numeric quantity in geometric (including astronomical event) and stochastic domains (such as card game poker hand probabilities)

We propose to utilize a new computing mode, multisite coordinated calculation, to simulate cultural evolution of fundamental mathematical concepts. That work is inherently interdisciplinary with computing and mathematics comprising a bit over half the effort, divided roughly equally between the former, essentially an engineering discipline, and the "queen of the sciences." The remainder of this effort involves historical, linguistic and art/architectural issues. Such matters constitute one of the key backgrounds of Allen Klinger and Nikos Salingaros. Such UCLA faculty as Jared Diamond and Peter Ladefoged in physiology and linguistics, respectively, have contributed to the knowledge of the author of this pre-proposal. Related references include [12-14]. Images play an important role in the evolution of human cultures from seeking to govern natural forces by magic to understanding them through science. See e.g., the image from [15] (also reproduced in [14]) [16]. But as [17] so convincingly demonstrates, computer-generation of designs like the [16] mandala-like instance, is an engineering reality today.

Finally, the issues we will cover are notably well-suited to multisite computing. UCLA is currently pursuing a patent following a disclosure by Allen Klinger and Byron Darrah. This involves innovations involving coding and software that enable cooperative and coordinated computing activities, and provides a means for enabling greater functionality in networked computers. We have new methods for incorporating multi-site computing in programs without requiring special administrative activity that would be particularly valuable in a multidisciplinary investigation. An instance of that is the issue of carbon-dating of human-settlement remains to determine domestication of food resources. For example, the Inca culture with only knots for number-records maintained a tax system over vast distances. Their food supply included the pepper known as chile manzana, the only domesticated variety known which does not now have a wild precursor. We will be able to bring multiply-skilled computer professionals into the process of investigating carbon-dating, number-system, and biological-relations among plants simultaneously using our multi-site computing methodology.

The effort would use multisite computing based on new computer software technology developed in the UCLA Computer Science Department in 1998 by Byron Darrah and Allen Klinger, with support from Bezu Arega, and fundamental mathematical concepts in pattern analysis, classification, and set theory. We would use direct knowledge by Mark Burgin of innovations by Mikhail M. Bongard recently discussed in the U.S. literature by Douglas Hofstadter.

The research would begin by developing four-person teams that include at least one computer-knowledgeable graduate student and two faculty. Selection of a cultural, historical, or physiological issue related to biosphere evolution or simply change is the first task for each team. The next stage involves establishing a simulation using multisite methods to replicate what could have occurred in situations where varied influences compete on a statistical basis. Finally a classification scheme would be constructed to model the eventual outcomes. That scheme and data yielded by the computational experiments would be presented to the entire participant group, first through weekly electronic exchanges, then through quarterly research meetings, and finally via circulation of a report publication.

[1] http://www.ucmp.berkeley.edu/history/linnaeus.html
[2] Bongard, Mikhail Moiseevich, Pattern Recognition (New York: Spartan Books, 1970); excerpts appear in http://www.stanford.edu/group/SHR/4-2/text/hofstadter.html, "on seeing A's and seeing As," by Douglas R. Hofstadter.
[3] Hofstadter, Douglas R., Godel, Escher, Bach : an eternal golden braid, NY: Vintage Books, 1980; also see the paper mentioned in [2].
[4] Burgin, Mark, Kuznetsov, Vladimir, Introduction to Modern Exact Methodology of Science, Moscow, ISF, 1994.
[5] Burgin, Mark, "Evaluation of Scientific Activity in Dynamic Information Theory," Science and Science of Science, 1994, No. 1-2, pp. 124-131; and Burgin, Mark, "Fundamental Structures of Knowledge and Information," Kiev, 1997.
[6] Klinger, Allen and Salingaros, Nikos, "A Pattern Measure," http://sphere.math.utsa.edu/sphere/salingar/PatternMeasure.html, Submitted for publication in Environment and Planning.
[7] Narins, Peter M., "Reduction of tympanic membrane displacement during vocalization of the arboreal frog, Eleutherodactylus coqui.," J. Acoust. Soc. Am., 91: 3551-3557, 1992.
[8] Diamond, Jared M., Guns, Germs and Steel: The Fates of Human Societies, NY: W. W. Norton & Company, 1997.
[9] Klinger, Allen, http://www.cs.ucla.edu/~klinger/modn.html and http://www.cs.ucla.edu/~klinger/newno.html (both describe an achievement by Peter Montgomery).
[10] Klinger, Allen, Zhen, Gu, and Stalnaker, Trask, http://www.cs.ucla.edu/~klinger/javadir/PatternMeasurer.html and http://sphere.math.utsa.edu/sphere/salingar/ .
[11] Burgin, Mark, "Triad as a Fundamental Structure in Human Culture," Studia Culturologica, 2, Spring 1993, 51-63.
[12] Klinger, Allen, "Recent Computer Science Research in Language Processing," Amer. J. of Computational Linguistics, 12: 3, 2-25, 1975.
[13] Joseph, George Gheverghese, The Crest of the Peacock - Non-European Roots of Mathematics, London UK: I.B. Tauris & Co. Ltd Publishers, 1991.
[14] Klinger, Allen, et. al., "What We Learned About p and Why We Looked at a Slowly-Converging Infinite Product P ," http://www.cs.ucla.edu/~klinger/pipaper.html
[15] Kasner, Edward, and James R. Newman, Mathematics and the Imagination, NY: Simon and Schuster, 1940; ISBN 1-55615-104-7 Redmond: WA, Tempus Books of Microsoft Press, 1989.
[16] Mandala, http://www.cs.ucla.edu/~klinger/radii.jpg
[17] Hofstadter, Douglas R., "on seeing A's and seeing As," http://www.stanford.edu/group/SHR/4-2/text/hofstadter.html (presents a modern computer-generated experimental set of alphabetic symbols).