The PDF format is used to allow ease of reuse of figures and text, quoted with citation.
Publication citations are given where a version of the report has been published. Often material has been edited in publication and the cited version differs to some extent.
A Stochastic Computer, Brian R Gaines, Internal Memo, Standard Telecommunication Laboratories, Harlow, UK, 1965. PDF.
Some notes on the application of digital circuits to the operations of arithmetic and differential calculus by means of a probabilistlc tepresentation of quantities.
Stochastic Computing Arrangement, Brian R Gaines, UK Patent 1,184,652, 1967. PDF.
The present invention relates to a form of computer which has particular application to computation with continuously variable quantities, for example in process control. According to the invention there is provided a computer including a digital cornputing element or elements, means for representing input information stochastically by the probability that a level in a clocked sequence of logic levels will be ON, means for applying the information so represented to the computing element or elements wherein computation is performed in a digital manner and means for converting the stochaslically represented outputs of the computing element or elements to analogue or digital values. The invention resides also in the method of representing quantities for computation and in apparatus for computation. simulation, pattern-recognition, model.formation, training and teaching. prediction. signal detection, automatic control, regulating reactions or for process control which uses the method of representing quantities or events which is herein described.
Stochastic computing, Brian R Gaines, Spring Joint Computer Conference, 30, 149-156, 1967. PDF.
The Stochastic Computer was developed as part of a program of research on the structure, realization and application of advanced automatic controllers in the form of Learning Machines. (A review of this article by Granino Korn is appended)
Stochastic computer thrives on noise, Brian R Gaines, Electronics, July 10, 1967, 72-79. PDF.
Many of today's computer control applications cannot be bandIed by conventional analog or digital machines, or even hybrid machines. A new approach, in which probability as an analog quantity switches a digital circuit, offers interesting promise. In this kind of machine, digital integrated circuits are randomly switched to simulate analog computer elements. Although slower than an analog machine and not as accurate as a digital one, the stochastic computer has a speed-size economy combination that cannot be matched by either. (A 1966 report in the same journal that is cited in this article is appended)
Techniques of identification with the stochastic computer, Brian R Gaines, Proceedings IFAC Symposium on "The Problems of Identification in Automatic Control Systems," Section 6 Special Identification Instruments, Prague June 12-19, 1967. PDF.
A major obstacle to the practical application of advanced techniques for process identification is lack of suitable hardware. By its very nature an identification computer must be able to store and adjust large numbers of variable parameters, and to use these for purposes of prediction and control. Conventional analog and digital computers both have disadvantages in this application, and there is a need for hardware specifically designed for identification purposes, economical in cost, reliable and drift-free like the digital computer, and capable of parallel operation to achieve the size-independent bandwidth of the analog computer. In particular this hardware should take full advantage of the advanced state of integrated circuit technology. This paper outlines the principles and structure of the Stochastic Computer, and describes three identification techniques of increasing generality which take especial advantage of novel features in stochastic computing elements. The first technique is one of steepest descent to the best linear relationship between the inputs and outputs of the process to be identified -- three conventional techniques, including polarity-coincidence correlation, are compared with three stochastic techniques. The second technique is one of statistical decision theory, in which Bayes Theorem is used to invert conditional probabilities and bring them to a form suitable for estimation and prediction. The final technique is one of Markov modelling of the state-class transitions of the process. In conclusion it is suggested that advances in integrated circuit technology mean that feasible control practice will, in a few years time, have advanced beyond control theory -- at least as it stands at present.
Phase computers, Brian R Gaines & P.L. Joyce, Proceedings of 5th Congress of International Association for Analog Computation, 48-57, Lausanne, 1967. PDF.
The phase computer is a modular, all.digital, dataprocessing and control system, in which many of the advantages of analog and digital computers are combined through the use of an incremental digital processor under program control. The incremental processor enables pseudo-analog computing loops to be established so that complex operations, such as division, squarerooting, rectangular/polar/hyperbolic co-ordinate transformations, and so on, are performed with the same speed and simplicity as simpler operations, such as addition and data transfer. The programming facility enables complex computations to be performed as a sequence of elementary operations, and hence overcomes the inordinate hardware demands of previous parallel incremental computers.
Varieties of computer—their applications and inter-relationships, Brian R Gaines, Keynote address at IFAC Symposium on Pulse Rate and Pulse Number Signals in Automatic Control, Budapest, 1968 PDF.
This paper discusses the impact of LSI on computer systems and computing techniques, and suggests that the major effects will be to make special-purpose computers for specific problem areas very much more attractive, and to blur the distinction between aoftware and hardware - in the sense that computer-assisted design and computer program compiling will merge into a general translation technique from problem specification to both .module interconnection (hardware) and variable stored-program (software). (Discussion is appended)
Foundations of stochastic computing systems, Brian R Gaines, Digest of IEEE International Convention. New York: IEEE, 1968. PDF.
During recent years considerable interest has arisen in the possibility of using random variables to represent quantities in a computer. This is an inefficient means of coding data but enables extremely simple digital hardware to be used to perform complex arithmetic functions. Stochastic computers, as systems using these representations are called, are capable of performing all the operations of the analog computer, addition, subtraction, multiplication, integration, and so on, using simple configurations of digital gates which are readily fabricated using LSI.
System organization, Brian R Gaines, Lecture Notes for IEE Vacation School on Engineering Aspects of Microelectronics, Department of Electrical Engineering Science, Colchester, UK, 1969. PDF.
Device technology has advanced rapidly during the last ten years, and made it possible to include in equipment data-processing sub-systems of increasing power and decreasing cost. However, this advance has not been uniform throughout the whole range of data-processing devices, neither has it proceeded uniformly in time. The trauma of transition from valve to transistor was over more rapidly than expected, but each successive change has had almost as much impact as that transition, and change follows change at a quickening pace. The transition from the early germanium transistors to silicon planar radically affected the speed of digital equipment, the accuracy of analogue equipment, and the temperature range of both. The coming of digital integrated circuits opened up a world of all-digital systems, and made the small digital computer a potential 'component'. However, integrated circuit operational amplifiers, have followed close behind, and their possibilities have been enhanced by thin-film networks and FET switches. Over the horizon looms LSI, and the very low-cost, micro-programmed digital machine, but we can be equally sure that over that same horizon are many developments of which we have taken no account, yet whose impact may be greater than that of LSI.
MINIC I Manual, Brian R Gaines, Department of Electrical Engineering Science, Colchester, UK, 1969. PDF.
MINIC is a modular, microprograrrmed minicomputer based on a set of functional blocks which may be plugged into standard hiway racking to form a range of specialised and general-purpose computers, controllers and data-camnunication systems. MINIC I consists of a set of MINIC modules which are microprograrrmed to behave as a general-purpose digital computer. This machine is byte orientated and the majority of instructions are 8 bits in length and operate on one or more 8-bit "bytes" of data.
Stochastic computing systems, Brian R Gaines, In J. Tou (Ed.), Advances in Information Systems Science, 2, 37-172, New York: Plenum Press, 1969. PDF.
This chapter reviews the data-processing requirements of pattern recognition and machine learning, and introduces the concept of stochastic computing through the representation of analog quantities by the probabilities of discrete events. The first part of the chapter gives a complete overview of stochastic computing and its relationship to other computational techniques. Later sections treat various forms of stochastic computer in some detail, and outline the theoretical basis for computing with probabilistic devices. The final sections describe various systems for pattern recognition and control where the use of stochastic computing elements is advantageous.
Stochastic computers, Brian R Gaines, In A. R. Meetham & R. A. Hudson (Eds.), Encyclopaedia of Linguistics, Information & Control, 66-76, London: Pergamon Press, 1969. PDF.
The stochastic computer was developed for the type of problem where the availability of large numbers of low-cost computing elements is more important than the speed and accuracy of computation. Lnformation within the computer is carried through modulation of the statistics of digital 'noise', and the theory underlying its computational processes is that of Markov chains and stochastic automata.
QUASIC Manual, Brian R Gaines, Department of Electrical Engineering Science, University of Essex, 1971. PDF.
QUASIC is a system and application programming language with a syntax based on that of Dartmouth College BASIC. It is the TSS8 (time shared PDP-8) version of the series of languages: BASYS for the PDP-9, MINSYS for the MINIC I and AIMS-11 for the PDP-11.
Real-time system design under an emulator embedded in a high-level language, Peter V Facey & Brian R Gaines, Proceedings DATAFAIR 73, 285-291, London: British Computer Society, 1973. PDF.
The paper discusses the problems of minicomputer software development and describes experience in the development of a small real-time computer system under an interactive emulator embedded in extended bASIC on the PDP10.
Computer technology and its utilization: today and tomorrow, Brian R Gaines, Proceedings of National Engineering Laboratory Conference on Small Computer Applications in Industry, East Kilbride, 1973. PDF.
The pace of advance in computer technology is so rapid that a major problem for industry is to maintain a comparable rate of advance in computer applications. If reliability and power continue to increase, and price, size and power consumption continue to fall, there is a strong incentive to sit on the fence and wait for a stable technology. However, will it come, when, and what will it be? In the meantime what strategies can be adopted to take advantages of minicomputer technology without suffering from the inherent obsolescence generated by rapid change.
MINSYS Manual, Brian R Gaines, Department of Electrical Engineering Science, University of Essex, 1971. PDF.
MINSYS is a system and application programming language with a syntax based on that of Dartmouth College BASIC. It is the MINIC I version of the series of languages: BASYS for the PDP-9, QUASIC for the TSS8 and AIMS-11 for the PDP-11.
An interactive display-based system for gilt-edged security broking, Brian R Gaines, Peter V. Facey & John Sams, European Computing Congress Conference Proceedings, EUROCOMP 74, 155-169, London: Online, 1974. PDF.
This paper describes experience in the development and use of a display-based interactive minicomputer system for gilt-edged security broking installed in the London Stock Exchange in January 1972. The system is located in the same office as the 12 dealers it serves and is continually up-dated with price information over a radio link from the exchange floor. For each security, current yields and deviations from the general trend are calculated and displayed on television monitors. Graphic terminals enable dealers to request individual stock-by-stock comparisons, fitted trend lines, and other facilities for investment analysis. The system is programmed in BASYS, a high-level language designed for ease of implementation of natural man-machine communication procedures.
Design objectives for a descriptor-organized minicomputer, Gaines, B. R., Facey, P. V., Williamson, F. K., & Maine, J. A., European Computing Congress Conference Proceedings, EUROCOMP 74,. 29-45, London: OnlineK, 1974. PDF.
This paper describes the logic and motivation behind the design of a descriptor-organised minicomputer for commercial production. It is argued that the availability of low-cost integrated circuit families has made it simple to achieve the conventional minicomputer design objectives (low-cost, fast-response, high reliability, and ease of interfacing), and that additional objectives are now both possible and necessary. The most attractive lines of development appear to be those related to simple, secure and swift software engineering, and. the requirements for the hardware support of process-structured operating systems and high-level languages are analysed. Finally a minicomputer design is outlined which combines microprogrammed multi-length operations, data descriptors and dynamic segmentation, linked to procedure calls, to satisfy frost of the detailed objectives established.
Integration of protection and procedures in a high-level minicomputer, Gaines, B. R., Haynes, M., & Hill, D., Proceedings IEE 1974 Computer Systems and Technology Conference. London: IEE, 1974. PDF.
This paper discusses those aspects of the design of a "high-level" minicomputer specifically orientated for ease and security of software development with high-level languages in a real-time environment. In particular an implementation of procedure calls is described which both supports the requirements of the common languages and integrates naturally with a ring-structured protection mechanism based on storage segmentation.
A high-level minicomputer, Williamson, F. K., Gaines, B. R., Maine, J. A., & Facey, P. V., Information Processing 74, 44-48, Amsterdam: Noth-Holland, 1974. PDF.
This paper describes the design considerations underlying the development of an advanced minicomputer (MINIC-S) now in commercial production. Emphasis is placed on programming/compiler and operating sytem requirements, on the one hand and engineering feasibility on the other. Microprogramming/trapping enables all machines of the range to offer the identical architecture and range of facilities. Descriptor-based data organisation enables a very wide range of operand types and lengths to be made available. Relocation/protection and a separate minicomputer I-O processor enable real-time process-structured operating systems to be implemented efficiently.
Some experience in interactive system development and application, Brian R Gaines & Peter V. Facey, Proceedings IEEE 63(6) June 1975, 894-911. PDF.
Minicomputers programmed in a high-level interactive language form a very attractive basis for the development of systems involving close man-computer collaboration. This paper is based on a wide range of experience of interactive minicomputer systems in commercial, medical, industrial, and scientific applications. It is first argued that the development of systems for effective man-computer collaboration requires not only interactive system use but also interactive system development. The designer needs to be able to tailor the system to user requirements at least partially as an experimental dialogue at a terminal with a user. There follows a critique of certain features of central computer utilities that limit their effectiveness in interactive applications, leading to a proposal for the use of minicomputer-based systems programmed, and used, interactively. We then give a number of case histories of our own experience in developing and using web systems in commercial, medical, and scientific applications From this experience, we have extracted a number of rubs for programming interaction between the user and computer system which we outlined. Finally, the main features of the software technology underlying these systems we briefly described.
The logic of protection, Ladislav J Kohout & Brian R Gaines, In G. Gods & J. Hartmanis (Eds.), GI-5. Jahrestagung, LNCS 34, 736-751, Berlin: Springer, 1975. PDF.
This paper presents a brief exposition of the role of various mathematical techniques in the development and utilization of resource protection structures for computers. The first section is concerned with the semantics of the problem - the distinction between protection problems in general and those whose complexity necessitates deeper theoretical treatment. The second section considers the roles of algebraic, topological, and modal/multi-valued logic, techniques in the analysis of protection. Finally we give an analysis of a current protection model to illustrate the problems and techniques.
Analogy categories, virtual machines, and structured programming, Brian R Gaines, In G. Gods & J. Hartmanis (Eds.), GI-5. Jahrestagung, LNCS 34, 691-700, Berlin: Springer, 1975. PDF.
This paper arises from a number of studies of machine/problem relationships, software development techniques, language and machine design. It develops a category-theoretic framework for the analysis of the relationships between programmer, virtual machine, and problem that are inherent in discussions of "ease of programming", "good programming techniques", "structured programming", and so on. The concept of "analogy" is introduced as an explicatum of the comprehensibility of the relationship between two systems. Analogy is given a formal definition in terms of a partially ordered structure of analogy categories whose minimal element is a "truth" or "proof" category. The theory is constructive and analogy relationships are computable between defined systems, or classes of system. Thus the structures developed may be used to study the relationships between programmer, problem, and virtual machine in practical situations.
Human factors in virtual machine hierarchies, Brian R Gaines, Proceedings Colloquium on the Influence of High Level Languages on Computer System Design, London: IEE, 1976. PDF.
These notes are a distillation of some basic considerations from diverse experiences, reported in detail elsewhere, in: minicomputer design, both for compact program encoding and for high level language support; language design for interactive systems; programming interactive dialogue; and analysing the foundations of structured programming, protection structures, and future trends in computers.
Protection as a general systems problem, Ladislav J Kohout & Brian R Gaines, International Journal of General Systems, 3(1), 3-23, 1976. PDF.
It is argued that the problem of protection, of controlling mutual access rights to shared resources, is a topic appropriately treated as a major component of general systems theory. Although most widely studied and developed in the context of computer systems, protection models arc equally applicable to biological systems, such as those involved in movement control. The paper first establishes the nature of the problem of protection in computer systems, noting that it only reaches its full potential complexity in large data-base systems with processes automatically invoked by "data interrupts". The Graham and Denning model of protection and the concept of a "capability" are then described and the appropriate mathematical tools for the analysis of such models discussed. A detailed model of protection is then developed with examples of the role of algebraic, automata-theoretic, topological and modal/multi-valued logical, techniques in its analysis. Finally, biological applications, general systems consequences and automatic design techniques, for protection structures are discussed.
Minicomputers in security dealing, Brian R Gaines, Peter V. Facey & John B. Sams, IEEE Computer, Vol. 9 No. 9, Sept. 1976, 6-15. PDF.
This paper describes one successful application of a minicomputer in a business environment where the machine and its terminals have been closely integrated into an existing office and provide both an improved means of distributing information and also on-line analysis of that information in a manner not previously possible. The emphasis is on the role of a computer system in a real financial dealing environment and the technical and operational features of the project which make it possible. In particular the results obtained back up our previous arguments [that the interactive systems programming language BASYS provides a foundation for rapid development of interactive minicomputer systems tailored in detail to user requirements and that the problems of systems specification and developmental modification of computer-based systems are greatly eased if the interactive nature of the system is fully utilized for on-line development in parallel with actual user experience.
Some notes on the provision of data acquisition, processing and retrieval in the small EDP environment, Brian R Gaines, Industry Report, Department of Electrical Engineering Science, University of Essex, Colchester, UK, 1976. PDF.
Whilst the conventional image of EDP computer is as a 'number cruncher' providing large-scale computational resources for a high throughput of numerical calculations, there are an increasing number of small EDP applications whose requirements are primarily those of accurate data acquisition, communication, storage and retrieval, rather than calculation. Even the classical database computations of search, sort and merge, are not dominant in these applications since the prime requirement is simple storage rather than global analysis.
Interpretive kernels for microcomputer software, Brian R Gaines, Proceedings Symposium Microprocessors at Work, 56–69, London: Society of Electronic & Radio Technicians, 1976. PDF.
The initial main applications area for microcomputers has been in small, mass~produced systems where they replace hardw1red, random logic. These present few problems of 'software' development because the 'programs' required are small and fixed. However, it is clear that the technology has now reached the state where the 'microcomputer' is in every sense a 'computer' with all the capabilities of much larger and more expensive machines. Increasingly many applications are looking towards its programmability, and continuing, in-use re-programmability. This is generating requirements for a level of software support not generally provided with microcomputers. More importantly it is forcing organizations with long experience of hardware manufacture to move into the area of software and systems development, maintenance and support. This paper is concerned with software engineering techniques that allow the same discipline of modularity, documentation, quality control, etc., that has previously been imposed on hardware to be applied to software.
Trends in stochastic computing, Brian R Gaines, Colloquium on Parallel Digital Computing Methods—DDA's and Stochastic Computing (Digest 1976/30). London: IEE, 19765. PDF.
Gaines, B. R. (1976). . The original development of the stochastic computer took place independently at Illinois and STL in 1965. Some 11 years later there is no dramatic progress to report and the 1969 technical survey is still adequate in most respects. However, the problem discussed therein, of the difficulty of simulating large systems (e.g. distillation columns) in real time, still remains. Meanwhile there has been steady progress in stochastio computing on several fronts.
BASYS—a language for processing interaction, Brian R Gaines & Peter V Facey, Proceedings Conference Computer Systems and Technology, IERE No. 36, 251-262, London: IERE, 1977. PDF.
Over the past 10 years we have had extensive experience in the development and application of low-cost, interactive, minicomputer-based systems designed for close collaboration between people and computers. In papers elsewhere we have described and analysed these systems and the human factors involved. In particular we have suggested strategies for programming the interaction between the naive user and the computer. This paper concentrates on the software technology underlying these systems and analyses in some detail the types of string handling facilities in particular that are needed to program interactive dialogue simply, flexibly and effectively. An outline is given of how these facilities are incorporated in the language BASYS that has been used in our applications.
AIMS-11 Manual, Peter V. Facey, Arbat Consulatnts, London, 1977. PDF.
AIMS is a system and application programming language with a syntax based on that of Dartmouth College BASIC. It is the PDP-11 version of the series of languages: BASYS for the PDP-9, QUASIC for the TSS8, and MINSYS for the MINIC I.
A mixed-code approach to commercial microcomputer applications, Brian R Gaines Conference on Microprocessors in Automation and Communications, IERE Conference Proceedings No.41, 291-301, London: IERE, 1979. PDF.
It has become very attractive to develop office automation systems based on microcomputers. However, currently no single one of the commonly available languages on such machines is suited to the full range of requirements: interactive dialogue; accounting; statistics; word-processing; data-base management: etc. This paper outlines the key requirements, tabulates the strengths and weaknesses of current languages, and describes a system that is now in use in a number of commercial applications which allows different parts of an application to be written in the most suitable language, yet the whole to be simply and uniformly integrated together.
Monotype cracks the Chinese puzzle, Clive Goodacre & Brian R Gaines Printing Today, 16(188), 54-60, 1979. PDF.
It taken Monotype only one year, following its rescue by the NEB, to deliver a production system that for almost a decade has beaten the resources of major companies throughout the world. In doing so it has swept away in one stroke all the Heath Robinson altempts at offering the Chinese an integrated photosetting system for coping with the mind-boggling complexity of their ideographic language. Which says a lot for a company that has been written off on numerous occasions as an anachronism left over from the letterpress era. At the same time as having to sort out the mess left over after the Grendon Trust asset stripping operation, Monotype has completed a production installation in Shanghai and Peking. This comprises 10 keyboards working on paper tape, 10 on floppy disc, four Lasercomp photosetters, two typographical digitisers and six editing termina Is each equipped with a Chinese character line printer. Additionally there are two modem communication units enabling any of the sub-systems to work over a telephone link.
The role of the behavioural sciences in programming, Brian R Gaines, Gaines, B. R. (1979). Structured Software Development Vol.2, 57-68, Maidenhead, UK: Infotech International, 1979. PDF.
This paper is about the role of the behavioural sciences in computer system implementation; about the contributions which one might reasonably expect from psychology and sociology rather than the actual contributions they have made. Indeed, from the outset, it should be made clear that the potential contribution by far exceeds any so far made and that the lessons to be learnt are not necessarily direct solutions to problems, but, perhaps more importantly, meta-systemic aspects of methodology and experimentation; that there are techniques of experimental investigation that may throw light on particular problems, but there are also approaches that most certainly will not. Both psychology and sociology are more rich in case histories of the experimenter misled, or the theorist misinterpreting experiments, than they are in generally useful 'laws'. We are on very dangerous ground when we attempt to draw general inferences about people's behaviour from observations of specific cases, even under closely controlled experimental conditions (perhaps, particularly under such conditions man is the most highly adaptive animal and changes his behavioural colouration to his local circumstances more rapidly than a chameleon changes its optical colouration to its local base!).
The bounds of possibility, Brian R Gaines Printing World, 16-17, 8th July 1981. PDF.
We now have digital typography. What is it? Why has it been developed? What can we do with it? Digital typography is simple. To specify the printed material a page is split uniformly both horizontally and vertically. This creates a matrix of minute picture elements - called pixels in computer jargon. Each pixel is either black or white. Any printed page may be specified in this way with complete accuracy using fine enough resolution. It is digital because the page is represented by a definite number of discrete elements each of which can take one of only two volumes. Call white the number 0 and black the number 1 and the picture on the page is just digitised data suitable for generation, storage and processing in a computer.
Unravelling the Chinese typesetting puzzle, Brian R Gaines In C. Goodacre (Ed.), Penrose 1981: International Review of the Graphic Arts, 25-40, London: Northwood Publications, 1981. PDF.
Chinese, the oldest recorded language in use today and spoken by one quarter of the world's population, is one of the few challenges left for modern computer photosetting systems. In September 1978 Monotype decided the time was ripe to tackle the problem and by December demonstrated an embryo system in Hong Kong. A joint development programme with the China Printing Corporation lead to a complete prepress system being developed and installed in Beijing and Shanghai by July 1979. After discussing the problems of Chinese typesetting, the author, who lead this remarkable project, describes existing solutions.
Some Experience in Text Processing in the Chinese Language, Brian R Gaines Technical Report, G W Information Transfer Systems, London, 1982. PDF.
The Chinese language present many difficulties in text processing. There are some 7,000 characters in routine use and conventional approaches to keyboards, displays and printers are unable to cope with the set required. Yet the language is a very important one since it is in daily use by one quarter of the population of the world. This paper describes a complete phototypesetting system recently developed for use with text in the Chinese and English languages and now in use for book printing in Beijing and Shanghai. Recent work on the application of a similar approach to data processing in Chinese is also outlined.
Digital know-how, Brian R Gaines Printing World Supplement, 13-14, June 1982. PDF.
At this Drupa we will see the release of many new products which were developed more than three years ago but whose production was deferred during the spending freeze of the recession. This reflects a general slowing down of applications of new technology in the print industry during this period. However, the underlying circuit, computer and communications technologies have continued their pace of development.
Office automation in the languages of the world, Brian R Gaines, Proceedings Graphics Interface ’8, 337-344, Toronto: Canadian Man-Computer Communications Society, 1983. PDF.
Computer-based information systems have been restricted to those languages using the Roman alphabet. In non-numeric applications, the limitation has led to an increasing differential in applicability between those countries using Latin languages and the majority which do not. Office automation, database and educational applications in particular have been severely restricted, and these are very important to the developing nations of the world. Low-cost graphics technology now provides the means for information systems to communicate as readily in "exotic" scripts as in Roman characters. Input, display and printing subsystems can be made available that cope with such scripts as Chinese, Devanagari, and Urdu Nastaliq. Problems remain in moving the operating systems, programming languages and application packages across language barriers. This paper is based on experience in developing systems for China, India and Pakistan, and covers both the technological ease of implemention and the remaining problems. It is concluded that Canada has a combination of computing and linguistic knowledge that would make development of "universallanguage" computer systems simple and rapid as a co-ordinated effort.
Origins of stochastic computing, Brian R Gaines, In W. J. Gross & V. C. Gaudet (Eds.) Stochastic Computing: Techniques and Applications, 13-37, Cham: Springer, 2019. PDF.
In the early 1960s research groups at the University of Illinois, USA, and Standard Telecommunication Laboratories (STL), UK, each independently conceived of a constructive use of random noise to implement analog computers in which the probability of a pulse in a digital pulse stream represented a continuous variable. The USA group initially termed this a noise computer but shortly adopted the UK terminology of stochastic computer. The target application of the USA group was visual pattern recognition, and that of the UK group was learning machines, and both developed trial hardware implementations. However, as they investigated applications they both came to recognize that the technology of their era did not support stochastic computing systems that could compete with avail- able computational technologies, and they moved on to develop other computing architectures, some of which derived from the stochastic computing concepts. Both groups published expositions of stochastic computing which provided a comprehensive account of the technology, the architecture of its functional modules, its potential applications and its then current limitations. These have become highly cited in recent years as new technologies and issues have made stochastic computing a competitive technology for a number of significant applications. This chapter provides a historical a historical analysis of the motivations of the pioneers and how they arrived at the notion of stochastic computing.
A conceptual framework for stochastic neuromorphic computing, Brian R Gaines, IEEE Design & Test, 38(6) 16-27, 2021. PDF.
This keynote article is written by Brian Gaines, the inventor of stochastic computing. He shares both a view back on the history of neuromorphic computing and a view forward on deep learning as a new information processing technology. Gaines observes that computing has been a recursive technology: it supports other technologies that in turn support the progress of computing itself, leading to a positive exponential feedback loop and an exponential growth. He infers that the same holds for deep learning with its ability to meta-learn solutions to its own design problems. (Ilia Polian, University of Stuttgart)
CPCS 6-Apr-2012