O.G. Selfridge. Pandemonium: A paradigm for learning. In Mechanisation of Thought Processes, pages 511-531. Her Majesty's Stationary Oce, 1959.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....[11] based on a weighted sum voting system, has been used to control mobile robots, both simulated and real, indoor and outdoor. For a theoretical discussion of the most popular voting systems, see [3] Voting has also been applied to pattern recognition tasks, the Pandemonium architecture [12] being an early example. Pandemonium is a layered classifier system in which each layer consists of demons that match certain features and shriek their outputs up to the next layer; the final output of the system depends upon whose shriek was the loudest. 3 Method As stated earlier, our goal ....

O. G. Selfridge. Pandemonium: A paradigm for learning. In Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, London: HMSO, November 1958.

....than a characteristic of the architecture. It is interesting to note that as early as 40 years ago, Selfridge regarded the notion of parallel processing as the basic motif behind his pandemonium model, probably one of the first team architectures for modeling human data processing techniques [14]. He gives two reasons: i) division of labor: it is often easier to handle data in a parallel manner , it is usually the more natural manner to handle it , ii) flexibility and adaptability: it is easier to modify an assembly of quasi independent modules than a machine all of whose parts ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes, pages 511 527. H.M.S.O., London, 1959.

....the importance of emotions for learning in an environmentally situated agent is discussed. The learning of complex emotions is dependent on Pandemonium Theory, which will be described first. Pandemonium Theory This architecture is based on a psychological theory called Pandemonium Theory (Selfridge 1959) who applied it only to perception. Later, John Jackson presented it to the computer science community in an extended and more concrete form (1987; Franklin 1995 ) that makes it useful for control of autonomous agents. In Jackson s version of Pandemonium Theory, the analogy of an arena is used. ....

Selfridge, O.G. 1959. Pandemonium: A Paradigm for Learning. In Proceedings of the Symposium on Mechanisation of Thought Process. National Physics Laboratory.

....not after. YMIR: 10 #An Architecture for Multimodal Agents An architecture is being developed, called mir, that directly addresses the above issues. Ymir is a hybrid system, based on the three layer model of multimodal dialogue, and incorporates features of a black board approach [Nii 1989, Selfridge 1959]: multiple knowledge sources cooperate to provide a solution to a problemin this case to interpret user actions and generate appropriate responses. Psychosocial expertise (dialogue management) is separated from the main interpretive process (content interpreter) in a modular fashion [Bolt, ....

Selfridge, O. G. (1959). Pandemonium: A Paradigm for learning. Proceedings of the Symposium on the Mechanization of Thought Processes, 511-529.

....enables subsequent learning mechanisms to be versatile and reliable [6] Specifically, a representation is useful for learning if it includes records of recurring and co occurring events. As noted by Barlow, a convenient substrate for such a representation is provided by Selfridge s Pandemonium [23]. In Barlow s Probabilistic Pandemonium, the response strength of a feature detector demon would be proportional to Gamma log P , where P is the probability of occurrence of the feature the demon detects. These signals are then propagated to an association network which receives unconditional ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London, 1959.

....desired intelligent behaviour indeed this depends very much on the environmental constraints and the network constraints in the form of the types of functionality available from the modules. The call for a step in the direction of the development of modular systems has frequently been made [Sel59, Min85, MP88b, Ede87, Mel88], but up till now most of the emphasis has been on developing small modular systems specialized as a whole for a particular task domain [Wai89, NKK90, SNA89, Ebe89] rather than systems capable, merely through a higher level alteration in the specification of the architecture or environment ....

....preprocessing modules to build up the image into sets of fundamental shapes. From here inputs could be provided to object classification recognition modules, which could further supply inputs for the classification of more complex objects composed of sets of the basic objects, as in Pandemonium I [Sel59]. The inputs and outputs of such modules are not however constrained to follow this specification nor is a module prevented a priori from receiving input from any other module (although one would in general desire a downward flow of information) In addition to classifying static objects, this ....

O. Selfridge. Pandemonium: a paradigm for learning. In The mechanisation of thought processes, H.M.S.O, 511--527, 1959.

....types of data) is connectionist theory [3] This allows parallel computations to take place with the results of each computation combining in some fashion to derive an overall conclusion. One of the earliest examples of this technique being applied to recognition was Selfridges pandemonium model [10] which was used for character recognition. One technique which has being developed [8, 9] to combine information sources is to use each source of information to instantiate all possible solutions that agree with the given view. The instantiated hypotheses are then either combined or assessed ....

O. G. Selfridge, "Pandemonium: A paradigm for learning," in Symposium on the Mechanization of Thought Processes; London, 1959.

....results of many small computations each looking at only a part of the input (Figure 1) Note that while the decision computation receives the results of the others, it, too, is local and limited in that it sees none of the system s environment. Related to the perceptron is the Pandemonium model [3], in which lower level demons shout their computational results to higher level demons. We shall apply genetic search to the problem of evolving perceptron structure, a process that itself forms an emergent computation. In Minsky and Papert s [2] terminology, the perceptron computes a global ....

....[4] introduced backpropagation. In this paper we investigate a quite different approach to the predicate modification problem, namely, genetic search over the space of possible perceptron structures. In proposing that the demons might mutate and undergo fusion to form new demons, Selfridge [3] made a related proposal for Pandemonium. Our approach will be to evolve a population of perceptrons under a Wilson 3 Perceptron Redux version of the genetic algorithm [5] in the context of a specific task. A general characteristic of the GA is that partial solutions to a problem, called schemata, ....

O. G. Selfridge, Pandemonium: a paradigm for learning, in: Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory (HMSO, London, 1958).

....camels, elephants, and bears. Each of the detectors for the familiar animals responds at a fraction of the maximal activity (the strength of the response is illustrated symbolically by the size of the disk beneath the detector box) Only part of the connections between layers are shown. Unlike in Selfridge s (1959) Pandemonium, all the responses and not merely the strongest one matter here. to defy an easy and general computational characterization. In face discrimination, for example, the physical variables best correlated with the principal dimensions identified by multidimensional scaling analysis of ....

Selfridge, O. G. (1959). Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London.

....of an open problem in constructive induction. 2.1 Recombination The recombination approach to constructive induction can be characterized as creating new attributes out of existing, or already known, attributes. It was first suggested by Selfridge, for use in his Pandemonium model of learning [Selfridge, 1959]; since then it has become the most commonly used approach to constructive induction. Recombination is popular because it requires very little information, it is easy to implement, and it is often sufficient. The major distinctions among different systems that do recombination are the operators ....

Selfridge, O. G. (1959). Pandemonium: A paradigm for learning. Proceedings of the Symposium on the Mechanization of Thought Processes (pp. 513-526). Teddington, England: National Physical Laboratory, H.M. Stationary Office, London.

....measure is the overall output weight of a unit. The overall output weight of a unit is the sum of the magnitudes of the weights on the connections between the output of the unit and other units. This measure was used, for example, by Samuel [119] in his checkers playing program, by Selfridge [123] in Pandemonium, and by Holland [59] in his bucket brigade algorithm. Other measures of worth include the product of the unit s output and its overall output weight and the cross correlation between the unit s output and the desired output. On comparing these three measures of worth in a two layer ....

Selfridge, O. Pandemonium: A paradigm for learning. In Proceedings of the Symposium on the Mechanisation of Thought Processes, Teddington, England: National Physical Laboratory, H.M. Stationary Office, London, 2 vols, 1959.

....that information is operated on sequentially by a single processor. Connectionism, instead, takes what we know about the brain as its basic foundation, resulting in a massively parallel, distributed information processing environment. Connectionist networks, in common with other models (e.g. Selfridge s Pandemonium system, 1959), use the activation of units, either singly or across a group of units, to represent hypotheses. These units, or nodes, interact via weighted connections which are often modifiable, allowing learning to occur. The reasons for the use of this substrate and its implications are examined in the ....

Selfridge, O. G. (1959). Pandemonium: A paradigm for learning. In Symposium on the Mechanisation of Thought Processes London: HMSO.

....and parity 10. Finally, in section 5 the performance of a Pandemonium system on optical character recognition (OCR) problems is discussed with reference to previous experimental results. 2 The Pandemonium system The Pandemonium architecture was originally proposed in 1958 by Selfridge [22]. The idea is to divide and conquer a complex problem domain, through use of a number of specialized agents working in parallel. All these agents, or daemons, process the same signal in parallel, and each provides a possible answer. The daemon that shouts the loudest is taken to be that which is ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....and discrimination. The need for module reliability is discussed, then improvements are proposed in the form of the Minos module, and its inclusion in an Authority system. Finally further extensions to the Authority system are briefly outlined. PANDEMONIUM I AND II In Selfridge s 1959 paper [15] he describes the Pandemonium system, which achieves efficiency in learning recognition problem domains through the utilization of several demons , each of which responds in a certain way to a particular input. Thus the input to a layer of demons would be the same for each, but the weightings ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....can be profitably used. After this point it we suggest that approximation can only be improved at a meta level of a modular system, whereby a number of different approximators (i.e. that generate significantly different approximation functions f) are used in a Pandemonium style parallel system [6, 9]. We further hold that most real world, large scale problems are of the OCR type, and that steps can most usefully be made in the direction of meta levels of approximation and reflective systems, rather than that of the details of the approximator itself. Acknowledgements I thank Uwe Beyer and ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....to build more complex aggregates, leading eventually to the categorization of the pattern. Many of these hierarchical models were statistical pattern recognition systems inspired by ideas from biology, including Rosenblatt s (1959) Perceptron, Uttley s (1959) Conditional Probability Computer, and Selfridge s (1959) Pandemonium. In the 1960s and 1970s a great deal of the research effort in computer vision went into the development of various edge finding schemes in order to extract reliable features to use as a starting point for object recognition and scene analysis (Clowes, 1971) Despite this effort, ....

Selfridge, O. (1959). Pandemonium: A paradigm for learning, Symposium on Mechanization of Thought Processes: National Physical Laboratory Symposium. London: HM Stationery Office.

....time, all of these features could be generated and tested but a more sophisticated control strategy should be developed, if only to increase the ratio of good to bad features. This generate and compete control strategy has been used in other AI systems, the earliest of which was probably Selfridge s (1959) Pandemonium system. Selfridge also talked about the worth of demons in Pandemonium, but this measure did not take into account the cost of their computations. Lenat s (1983) AM system for mathematical discovery maintained two separate rating systems. Each concept had an interestingness value, ....

Selfridge, O. G. (1959). Pandemonium: A paradigm for learning. Proceedings of the Symposium on the Mechanization of Thought Processes (pp. 513-526). Teddington, England: National Physical Laboratory, H.M. Stationary Office, London.

....experimenter. For example, if a number of experimental setups is possible the experimenter would be able to interact with JANUS to find a configuration that JANUS likes best. The second area is concerned with internally directed reflection. This is employed in the construction of reagent teams [2, 19, 18]. These structures allow the choice of one of many possible models for solution offered by team members. For such environments confidence is required to question answer confidence reagent outside world inside system Figure 5: Reflection as a means to communicating with the external world. allow ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....are different methods for solving a problem. Each of these methods has its advantages and disadvantages. In a team the problem is presented before a set of team members. Each member uses a different method of solving the problem, very much like in Selfridge s Pandemonium system and its derivatives [14, 15]. A 0 A 1 A i s g m c Master m c m c m c A c g G Figure 3: The idea of a team architecture Since a team output only one answer at a time the answers of the different team members must be combined by a team master. This integration process is one of the main problems in a teamlike ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....of a set of classifiers. The problem of combining classifier systems has received a large amount of recent attention (e.g. Wolpert, 1993 ] but has been a recognized area of research for quite a long time. In seminal artificial intelligence research such as Selfridge s Pandemonium [ Selfridge, 1959 ] and Nilsson s committee machines [ Nilsson, 1990 ] originally published in 1965) the idea of combining classifiers was advanced. In 1989, Clement reviewed over 200 papers on the more general issue of combining forecasts [ Clement, 1989 ] Particular research interest 24 recently has been ....

....classifiers, and so on. The idea of a layered classifier is not original to Wolpert, however, and can be traced at least as far back as Selfridge s Pandemonium for pattern recognition, whose architecture 28 is very similar to the standard two layer stacked generalizer as shown in Figure 1. 1 [ Selfridge, 1959 ] The idea of stacked generalization is quite simple. The algorithm assumes that we have been given a set of level 0 classifiers, a level 1 classifier, and a data set of classified examples with real valued features. The classes are real values as well under Wolpert s assumed framework. The ....

[Article contains additional citation context not shown here]

Selfridge, O. G. 1959. Pandemonium: A Paradigm for Learning. In Proceedings of the Symposium on the Mechanization of Thought Processes, Teddington, England. National Physical Laboratory, H.M. Stationery Office, London. 511--529.

....Multi agent systems, distributed machine learning, organizational design. 1. Introduction The concept of multi agent systems has a long tradition in computer science and artificial intelligence. It has influenced, for instance, the initial developments in areas such as cognitive modelling [36, 42], blackboard systems [12] object oriented programming languages [22] and formal models of concurrency [8, 40] Nowadays this concept gives new perspectives in computer science [6, 30] and establishes a major research and application field in distributed artificial intelligence (e.g. 5, 7, 10, ....

O.G. Selfridge, Pandemonium: a paradigm for learning, in: Proceedings Symposium on Mechanisation of Thought Processes , London, England (1959) 511--529.

....of Decision Making Dominic W. Massaro The present research tests a viable model of pattern recognition as a description of cognitive probability judgments. The model, called a fuzzy logical model of perception (FLMP) formalizes developments in fuzzy logic (Zadeh, 1965) pattern recognition (Selfridge, 1959), and choice theory (Luce, 1959) to provide a systematic account of perceptual judgments. The FLMP has been tested extensively in a wide variety of domains and provides an informative account of the fundamental processes involved in situations in which there are multiple sources of information ....

Selfridge, O. G. (1959). Pandemonium: A paradigm for learning. In Mechanization of thought processes (pp. 511-526). London: Her Majesty's Stationery Office.

....regions of specialization results in the modular decomposition recomposition problem (section 4. We return to it once more with respect to the decomposition afforded in the Pandemonium system in section 13. 3 Pandemonium The Pandemonium architecture was originally proposed in 1958 by Selfridge [26]. The idea is basically one of dividing and conquering a complex problem domain, through use of a number of specialized agents working in parallel. All these agents, or demons, process the same signal, or question, in parallel, and each provides a possible answer. The demon that shouts the ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

....to the notion of behavior releasing mechanisms, borrowed from ethology (Barlow, 1979) Its generalization to higher perceptual functions such as shape recognition was subsequently attempted. A well known proposal for an object recognition scheme based on feature detectors the Pandemonium (Selfridge, 1959; Lindsay and Norman, 1977) consisted of a three level hierarchy: feature demons (responsible for the detection of lines, corners, etc. cognitive demons (responsible for entire objects) and a master demon (responsible for the recognition decision) The limited influence of the Pandemonium ....

Selfridge, O. G. (1959). Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London.

....to the notion of behavior releasing mechanisms, borrowed from ethology (Barlow, 1979) Its generalization to higher perceptual functions such as shape recognition was subsequently attempted. A well known proposal for an object recognition scheme based on feature detectors the Pandemonium (Selfridge, 1959; Lindsay and Norman, 1977) consisted of a three level hierarchy: feature demons (responsible for the detection of lines, corners, etc. cognitive demons (responsible for entire objects) and a master demon (responsible for the recognition decision) The limited influence of the Pandemonium ....

Selfridge, O. G. (1959). Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London.

....standalone basis, however. The idea can be traced at least as far back as Nilsson s committee machines [1990] originally published in 1965) and Selfridge s Pandemonium for pattern recognition, whose architecture is very similar to the standard two layer stacked generalizer as shown in Figure 1. 1 [Selfridge, 1959] . Pandemonium was a winner take all system in which only the prediction of the agent with the loudest shout (the greatest level of confidence) was applied. Pandemonium did not attempt to combine the predictions of the pattern recognizing agents, however. Stacked generalization is a framework for ....

Selfridge, O. G. 1959. Pandemonium: A Paradigm for Learning. In Proceedings of the Symposium on the Mechanization of Thought Processes, Teddington, England. National Physical Laboratory, H.M. Stationery Office, London. 511--529.

....(MDS) work: in a metric space, fixing the relative distances of a set of 3 Webster s Dictionary has: cho.rus n ko r s, ko.r n n [L, ring dance, chorus, fr. Gk choros] 1a: a company of singers and dancers in Athenian drama participating in or commenting on the action. In a chorus, unlike in Selfridge s (1959) Pandemonium, the contributions of the individual actors are in harmony with each other. Lee Brooks (1987, p.165) uses the expression chorus of instances in his discussion of Medin and Schaffer s (1978) theory of representation. 4 But which may also correspond to entities devoid of a real ....

....camels, elephants, and bears. Each of the detectors for the familiar animals responds at a fraction of the maximal activity (the strength of the response is illustrated symbolically by the size of the disk beneath the detector box) Only part of the connections between layers are shown. Unlike in Selfridge s (1959) Pandemonium, all the responses and not merely the strongest one matter here. points effectively determines their coordinates, up to a translation and rotation of axes (Shepard, 1980) If several basic requirements, listed in section 4.1 below, are satisfied, and if the input stimuli do in fact ....

Selfridge, O. G. (1959). Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London.

....agents. Systems partitioned into groups of experts feeding single decision elements are at least as old as the field of AI itself. In an 1959 conference submission, Oliver Selfridge proposed a system for pattern recognition and learning anticipating elements crucial to much of today s work [10]. Multiple cognitive demons independently inspect the input data and submit their analysis to the decision demon in a bid for dominance. In the context of 3D image understanding, Adelson and Pentland [1] suggest an approach for disambiguating changes in surface reflectance from illumination ....

Oliver G. Selfridge. Pandemonium : A paradigm for learning. In The Mechanisation of Thought Processes. H.M. Stationary Office, London, 1959.

....et al. 1959) and was linked to the notion of behavior releasing mechanism, borrowed from ethology. It was quickly generalized to encompass higher perceptual functions such as shape recognition. A well known proposal for an object recognition scheme based on feature detectors the Pandemonium (Selfridge, 1959; Lindsay and Norman, 1977) consisted of a three level hierarchy: feature demons (responsible for the detection of lines, corners, etc. cognitive demons (responsible for entire objects) and a master demon (responsible for the recognition decision; cf. Figure 8) The proponents of the ....

Selfridge, O. G. (1959). Pandemonium: a paradigm for learning. In The mechanisation of thought processes. H.M.S.O., London.

....two spirals and parity 10. Finally, in section 5 the performance of a Pandemonium system on optical character recognition (OCR) problems is discussed with reference to previous experimental results. 2 The Pandemonium system The Pandemonium architecture was originally proposed in 1958 by Selfridge [24]. The idea is to divide and conquer a complex problem domain, through use of a number of specialized agents working in parallel. All these agents, or daemons, process the same signal in parallel, and each provides a possible answer. The daemon that shouts the loudest is taken to be that which ....

O. G. Selfridge. Pandemonium: a paradigm for learning. In The Mechanisation of Thought Processes: Proceedings of a Symposium Held at the National Physical Laboratory, November 1958, pages 511--527, London: HMSO, 1958.

.... has been directed at devising methods that can learn or build higher order feature detectors automatically; examples of this approach to vision are the work of Linsker, 1988, the TRAFFIC system (Zemel et al. 1988) and, from the early days of neural network research, the Pandemonium model (Selfridge, 1959). While, ultimately, adaptive systems like neural networks are almost certainly needed, most neural network models are designed as black boxes , without a detailed understanding of the combinatorial and statistical structure and constraints of the vision problem. Therefore, the approach described ....

Selfridge O. G., 1959, Pandemonium: A paradigm for learning., In The Mechanisation of Thought Processes, London: H.M. Stationary Office.

....by electrical, chemical, or mechanical means. Furthermore, signals could be supplied from an external stimulus (such as light impinging on the retina) or from other processing units. The processing units (see Figure 1) may refer to neurons, mathematical functions, or even demons a la Selfridge [100]. Lastly, information may be encoded in the units either locally or in a distributed manner. S Inputs Processing Unit Outputs 1 1 e input i S ( i ) q (a) b) Figure 1: Different forms of processing units: a) stylized sympathetic ganglion, b) mathematical function. Connections between ....

....of signals [66] Finally, the weights associated with the connections may be hardwired , learned, or both. The weights represent the strength of connection (either excitatory or inhibitory) between two units. These three tenets allow a large spectrum of models (e.g. Selfridge s Pandemonium [100]; Rumelhart McClelland s PastTense Acquisition Model [95] Dawson s Motion Correspondence Model [18] to fall within the classification of connectionist research. To understand how these different models fit into connectionist research today, one needs to be aware of how connectionist research ....

[Article contains additional citation context not shown here]

O. G. Selfridge. Pandemonium: A paradigm for learning. In D. V. Blake and A. M. Uttley, editors, Proceedings of the Symposium on Mechanisation of Thought Processes, pages 511--529, London, 1959. H. M. Stationary Office.

....a constructivist approach to intelligence. Divide and conquer is a natural approach for complex computational tasks. Thus here too one finds a growing consensus that human caliber intelligence must somehow originate in the interaction of many simpler agents c.f. 17] 24] 31] 35] 39] 44] [50], 56] 57] An extended abstract of this paper appeared in ICML 96[6] Arguably the critical problem in producing intelligence as a multiagent system is attributing credit when several agents together collaborate on a project. As Minsky wrote already in 1963[40] suppose that one million ....

....and incremental payoff, Hayek generated a general and efficient solution to arbitrary size problems the first time it evidently could express one. 5 Related Work The origin of mental capabilities in the interaction of smaller units has been widely studied, e.g. 17] 24] 31] 35] 39] 44] [50], 57] 56] These models are not explicitly economic. We commented in the introduction that noneconomic multi agent models risk unintended consequences arising from distorted implicit incentives. For one example, Miller and Drexler [38] discuss how imprecise motivation of meta agents led to ....

Selfridge, O. G. (1959) "Pandemonium: a paradigm for learning", Proceedings of the Symposium on Mechanisation of Thought Process. National Physics Laboratory.

....a number of important details. For instance: Is the lexicon passive or active In some models, the lexicon is a passive data structure (Forster, 1976) In other models, lexical items are active (MarslenWilson, 1980; McClelland Elman, 1986; Morton, 1979) in the style of Selfridge s demons (Selfridge, 1958). How is the lexicon organized and what are its entry points In active models, the internal organization of the lexicon is less an issue, because the lexicon is also usually content addressable, so that there is direct and simultaneous contact between an unknown input and all relevant lexical ....

Selfridge, O.G. (1958). Pandemonium: A paradigm for learning. Mechanisation of thought processes: Proceedings of a symposium held at the National Physical Laboratory, November 1958. London: HMSO.

No context found.

O.G. Selfridge. Pandemonium: A paradigm for learning. In Mechanisation of Thought Processes, pages 511-531. Her Majesty's Stationary Oce, 1959.

No context found.

Selfridge, O.G.: Pandemonium: A paradigm for learning. In Blake, D., Uttley, A., eds.: The Mechanisation of Thought Processes. Volume 10 of National Physical Laboratory Symposia. Her Majesty's Stationary O#ce, London (1959) 511--529

No context found.

Selfridge, O.G.: Pandemonium: A paradigm for learning. In Blake, D., Uttley, A., eds.: The Mechanisation of Thought Processes. Volume 10 of National Physical Laboratory Symposia. Her Majesty's Stationary O#ce, London (1959) 511--529

No context found.

O.G.Selfridge. Pandemonium: A Paradigm for Learning. In Mechanisation of ThoughtProcesses: Proceedings of a symposium held at the National Physical Laboratory, November 1958.

No context found.

Selfridge, O. (1958) `Pandemonium: A Paradigm for Learning,' Symposium on the Mechanization of Thought. London: HM Stationery Office.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC