31 March 2012

Computational Thinking: Some thoughts about Abduction

One of the striking features of computation is the extent to which forms of pattern matching are required in computer processing. Pattern recognition can be described as a means of identifying repeated shapes or structures which are features of a system under investigation. Whilst we tend to think of patterns as visual, of course they can also be conceptual, iterative, representational, logical, mathematical, etc. in form providing the underlying computational system can be programmed to recognise the distinctive shape of the pattern from the data. They can also consist of meta-patterns as described by Gregory Bateson as patterns that can detected across different spheres, such as culture, humanities, science and the social or 'the pattern that connects' (see Bateson 1979; Dixon 2012). The recognition of patterns and uncovering their relationships in sets of data was called 'abductive reasoning' by Charles Peirce, who contrasted it with inductive and deductive reasoning. Indeed, Peirce described abduction as a kind of logical inference akin to guessing. This he called the leap of abduction where by one could abduce A from B if A is sufficient (or nearly sufficient) but not necessary for B. The possible uses of this within a computational context should be fairly obvious, especially when software is handling partial, fuzzy or incomplete data and needs to generate future probabilistic decision points, or recognise important features or contours in a data set.

Charles Sanders Peirce (1839–1914) argued that pattern matching, which he called abduction or retroduction (he also used the terms presumption or hypothesis), was a type of hypothesis formation. The crucial function of 'a pattern of abduction … consists in its function as a search strategy which leads us, for a given kind of scenario, in a reasonable time to a most promising explanatory conjecture which is then subject to further test' (Schurz 2008, 205). Peirce argued,
Abduction is the process of forming an explanatory hypothesis.  It is the only logical operation which introduces any new idea; for induction does nothing but determine a value, and deduction merely evolves the necessary consequences of a pure hypothesis. Deduction proves that something must be; Induction shows that something actually is operative; Abduction merely suggests that something may be (Pearce 1958: 5.171, original emphasis).
Or perhaps better:
The abductive suggestion comes to us like a flash. It is an act of insight, although extremely fallible insight. It is true that the different elements of the hypothesis were in our minds before; but it is the idea of putting together what we had never before dreamed of putting together which flashes the new suggestion before our contemplation (Pearce 1988: 227, original emphasis).
In effect, abduction is the process of arriving at an explanatory hypothesis or a process of generating a hypothesis. As Eldridge explains,
For Peirce, abduction works from these surprising facts to determine a possible, plausible explanation.  Furthermore, Peirce stresses the fact that the logic of abduction is fallible – abductive inferences, like induction, can, and do, lead us to the wrong result (Pearce 1958 5.189, 5.197, 6.532).  However, as a part of the triad, abduction is able to correct itself, once it is investigated by deduction and tested by induction (Pearce 1958 5.574).  Because of this, we should never take the conclusion of an abductive inference to be a fact in and of itself until it is tested.  Until that point “abduction commits us to nothing…it merely causes a hypothesis to be set down upon our docket of cases to be tried” (Pearce 1958 5.602).  Furthermore, by hypothesis, Peirce does not just mean scientific hypotheses.  Abduction certainly includes the more formalized, conscious cognitive process of deliberately searching for an explanation to a set of particular facts; however, abduction is also a logical inference used in everyday life from crude hypotheses (his Catholic priest example) to perceptual judgments (understanding the information that we receive from our senses) (Pearce 1958 7.202, 5.180, 5.184) (Eldridge n.d.). 
Patterns were made popular as a heuristic for thinking about the new problematics introduced by software systems through the work of the architect Christopher Alexander (1936-), particularly Notes on the Synthesis of Form (Alexander 1964), The Timeless Way of Building (Alexander 1979), and A Pattern Language (Alexander et al. 1977) which influenced computer scientists, who found useful parallels between building design and the practice of software design (Rybczynski 2009)Alexander's central premise in his books, 'driving over thirty years of thoughts, actions, and writings, is that there is something fundamentally wrong with twentieth century architectural design methods and practices' (Lea 1997). Indeed, A Pattern Language was originally written to enable any citizen to design and construct their own home although it is arguable that he has had more influence on computer scientists than architects. As Appleton explains, patterns 'are a literary form of software engineering problem-solving [approach] that has its roots in a design movement of the same name in contemporary architecture... [they enable a] common vocabulary for expressing its concepts, and a language for relating them together. The goal of patterns within the software community is to create a body of literature to help software developers resolve recurring problems encountered throughout all of software development' (Appleton 2000).
The Timeless Way of Building and A Pattern Language were written as a pair, with the former presenting rationale and method, and the latter concrete details. They present a fresh alternative to the use of standardized models and components, and accentuate the philosophical, technical and social-impact differences between analytic methods and the adaptive, open, and reflective (all in several senses) approach to design. The term pattern is a preformal construct (Alexander does not ever provide a formal definition) describing sets of forces in the world and relations among them. In Timeless, Alexander describes common, sometimes even universal patterns of space, of events, of human existence, ranging across all levels of granularity. A Pattern Language contains 253 pattern entries. Each entry might be seen as an in-the-small handbook on a common, concrete architectural domain. Each entry links a set of forces, a configuration or family of artifacts, and a process for constructing a particular realization. Entries intertwine these 'problem space', 'solution space', and 'construction space' issues in a simple, down-to-earth fashion, so that each may evolve concurrently when patterns are used in development (Lea 1997).
Patterns are therefore reusable, structured, or formalised ways of doing things or processing information and data. Alexander himself defined each pattern as:

a three-part rule, which expresses a relation between a certain context, a problem, and a solution. As an element in the world, each pattern is a relationship between a certain context, a certain system of forces which occurs repeatedly in that context, and a certain spatial configuration which allows these forces to resolve themselves. As an element of language, a pattern is an instruction, which shows how this spatial configuration can be used, over and over again, to resolve the given system of forces, wherever the context makes it relevant. The pattern is, in short, at the same time a thing, which happens in the world, and the rule which tells us how to create that thing, and when we must create it. It is both a process and a thing; both a description of a thing which is alive, and a description of the process which will generate that thing (Alexander 1979: 247).
The antithesis to a pattern, is called an anti-pattern, that is patterns that describe (i) a bad solution to a problem which resulted in a bad situation, or (ii) describe how to get out of a bad situation and how to proceed from there to a good solution (Appleton 2000; Brown et al. 1998).  Patterns and pattern languages provide a broader framework to think about questions of paradigmatic means of designing and implementing computational systems. Indeed, in many cases patterns are used in this way to indicate a set of means for the development of software at a macro level. It should also be noted that patterns can be combined with other patterns to produce new patterns at a higher level of complexity, indeed this is the idea behind Alexander's (1977) notion of a 'pattern language'. Within software design it is quite common to see three levels noted, namely from most abstract to more concrete: Architectural Patterns, Design Patterns and Implementation Patterns, the last being detailed, programming-language-specific patterns as idioms (Microsoft 2012).

Within computer science, and particularly related to the more micro level problem of recognising patterns themselves within data sets automatically using computation, is an important and challenging area of research. The main forms of pattern recognition (we can think of these as patterns to find patterns) used in computation are usually enumerated as template-matching, prototype matching, feature analysis, recognition by components, fourier analysis, and lastly bottom-up and top-down processing. I'll briefly describe each of the six main approaches.

Template Matching: This is where a computational device uses a set of images (or templates) against which it can compare a data set, which might be an image for example (for examples of an image set, see Cole et al. 2004).

Template Matching (Jahangir 2008)
Prototype Matching: This form of patten matching uses a set of prototypes, which are understood as an average characteristic of a particular object or form. The key is that there does not need to be a perfect match merely a high probability of likelihood that the object and prototype are similar (for an example, see Antonina et al. 2003).

Feature Analysis: In this approach a variety of approaches are combined including detection, pattern dissection, feature comparison, and recognition. Essentially the source data is broken into key features or patterns to be compared with a library of partial objects to be matched with (for examples, see Morgan n.d.).

Recognition by Components: In this approach objects are understood to be made up of what are called 'geons' or geometric primitives. A sample of data or images is then processed through feature detectors which are programmed to look for curves, edges, etc. or through a geo detector which looks for simple 2D or 3D forms such as cylinders, bricks, wedges, cones, circles, and rectangles (see Biederman 1987).

Fourier Analysis: This form of pattern matching uses algorithms to decompose something into smaller pieces which can then be selectively analysed. This decomposition process itself is called the Fourier transform.  For example, an image might be broken down into a set of twenty squares across the image field, each of which being smaller, is made faster to process. As Moler (2004) argues, 'we all use Fourier analysis every day without even knowing it. Cell phones, disc drives, DVDs, and JPEGs all involve fast finite Fourier transforms'. Fourier transformation is also used to generate a compact representation of a signal. For example, JPEG compression uses a variant of the Fourier transformation (discrete cosine transform) of small square pieces of the digital image. The Fourier components of each square are then rounded to lower arithmetic precision, and weak components are discarded, so that the remaining components can be stored in much less computer memory or storage space. To reconstruct the image, each image square is reassembled from the preserved approximate Fourier-transformed components, which are then inverse-transformed to produce an approximation of the original image, this is why the image can produce 'blocky' or the distinctive digital artefacts in the rendered image, see JPEG (2012).

Bottom-up and Top-down Processing: Finally, in the Bottom-up and Top-down methods an interpretation emerges from the data, this is called data-driven or bottom-up processing. Here the interpretation of a data set to be determined mostly by information collected, not by your prior models or structures being fitted to the data, hence this approach looks for repeated patterns that emerge from the data. The idea is that starting with no knowledge the software is able to learn to draw generalisations from particular examples. Alternatively an approach where prior knowledge or structures are applied data is fitted into these models to see if there is a 'fit'. This approach is sometimes called schema-driven or top-down processing. A schema is a pattern formed earlier in a data set or drawn from previous information (Dewey 2011).

What should be apparent from this brief discussion of the principles of abduction and pattern-matching in computer science is their creative possibilities for generating results from data sets. The ability to generate  hypothesises on the basis of data, which is fallible and probabilistic allows for computational devices to generate forecasts and predictions based on current and past behaviours, data collection, models, and images. It is this principle of abductive reason which makes computational reasoning different from instrumental reason, and particularly from the iron-cage of logical implication or programmatic outcome that instrumental reason suggests. Indeed Alexander that the most useful patterns are generative,
These patterns in our minds are, more or less, mental images of the patterns in the world: they are abstract representations of the very morphological rules which define the patterns in the world. However, in one respect they are very different. The patterns in the world merely exist. But the same patterns in our minds are dynamic. They have force. They are generative. They tell us what to do; they tell us how we shall, or may, generate them; and they tell us too, that under certain circumstances, we must create them. Each pattern is a rule which describes what you have to do to generate the entity which it defines. (Alexander 1979: 181-182)




Bibliography

Alexander, C. (1964) Notes on the Synthesis of Form, Harvard University Press.

Alexander, C., S. Ishikawa, & M. Silverstein (1977) A Pattern Language, Oxford: Oxford University Press.

Alexander, C. (1979) The Timeless Way of Building, Oxford: Oxford University Press.

Antonina, K., Barbro, B., Hannu, V., Jarmo, t. and Ari, V. (2003) Prototype-Matching System for Allocating Conference Papers, accessed 31/03/2012, http://www.hicss.hawaii.edu/HICSS36/HICSSpapers/DTDMI06.pdf

Appleton, B. (2000) Patterns and Software: Essential Concepts and Terminology, accessed 31/03/2012, http://www.cmcrossroads.com/bradapp/docs/patterns-intro.html

Bateson, G. (1979) Mind and Nature: A Necessary Unity, (Advances in Systems Theory, Complexity, and the Human Sciences). Hampton Press, accessed 31/03/2012, http://www.oikos.org/mind&nature.htm

Biederman, I. (1987) Recognition-by-Components: A Theory of Human Image Understanding, Psychological Review, 1987, Vol. 94, No. 2,115-147, accessed 31/03/2012, http://www.cim.mcgill.ca/~siddiqi/COMP-558-2012/rbc.pdf

Brown, W., Malveau, R., McCormick, H. and Mowbray, T. (1998) AntiPatterns, accessed 31/03/2012, http://www.antipatterns.com/

Cole, L, Austin, D., Cole, L. (2004) Visual Object Recognition using Template Matching, accessed 31/03/2012, http://www.araa.asn.au/acra/acra2004/papers/cole.pdf

Dewey, R. A. (2011) Top-down and Bottom-up Processing http://www.intropsych.com/ch07_cognition/top-down_and_bottom-up_processing.html

Dixon, D. (2012) Analysis Tool or Design Methodology? Is there an epistemological basis for patterns?, in Berry, D. M. (ed.) Understanding Digital Humanities, London: Palgrave.

Eldridge, M. (n.d.) Clarifying the Process of Abduction and Understanding “Inductive” Generalization, accessed 31/03/2012, http://www.philosophy.uncc.edu/mleldrid/SAAP/USC/TP26.html

Janhangir, N. (2008) Genetic Algorithm Driven Template Matching In ActionScript 3.0, accessed 31/03/2012, http://nadimissimple.wordpress.com/2008/12/11/genetic-algorithm-driven-template-matching/

JPEG (2012) JPEG Homepage, accessed 31/03/2012, http://www.jpeg.org/jpeg/index.html

Lea, D. (1977) Christopher Alexander: An Introduction for Object-Oriented Designers, accessed 31/03/2012, http://g.oswego.edu/dl/ca/ca/ca.html

Microsoft (2012) Organizing Patterns, accessed 01/04/2012, http://msdn.microsoft.com/en-us/library/ff647589.aspx

Moler, C. (2004) Numerical Computing with MATLAB, accessed 31/03/2012, http://www.mathworks.se/moler/chapters.html

Morgan, M. (n.d.) Feature Analysis, accessed 31/03/2012, http://www.staff.city.ac.uk/~morgan/FeatureAnalysis.pdf

Peirce, C. S. (1958) The Collected Works of Charles Sanders Peirce, Harvard University Press.

Peirce, C. S. (1988) Pragmatism as the Logic of Abduction, in The Essential Peirce: Selected Philosophical Writings, 1893—1913, Bloomington: Indiana University Press.

Rybczynski, W. (2009) Do You See a Pattern?, Slate, accessed 31/03/2012, http://www.slate.com/articles/arts/architecture/2009/12/do_you_see_a_pattern.html

Schurz, G. (2008) Patterns of Abduction, Synthese, 164 (2): 201-234.

No comments:

Post a Comment

Disqus for Stunlaw: A critical review of politics, arts and technology