APL94 Abstracts

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 7

APL94 will be held in Antwerp, BELGIUM, 11-15 September 1994¶

APL — the Language and its Applications¶

Opening session: How symbols and notations may affect the mind and brain Dr. Paul J. Lewi, Janssen Research Foundation, Beerse, Belgium. Computer science produces powerful symbols and notations. Psychoanalysts proposed a model for the mind. Neurophysiology begins to reveal the molecular structure of the brain. The brain appears as a gigantic parallel computer that continuously rewires itself: plastic hardware. Patterns, when reinforced, grow into synaptic connections between neurons: sticky software. Symbols are the first links in the chain of encoding that maps the world into our brains. They mediate between the observer and the observed. Primitive people and alchemists understood the charm of symbols. Modernist artists have rediscovered their magic. Present-day scientists are regaining their alliance, most recently in fractal geometry and chaos theory. An understanding of symbols and notations is essential for our intellectual, if not mental, health. Because they embody our hopes, beliefs and fears, they record the progression and fate of mankind.

Plenary Presentations¶

APL in Finance Arthur Whitney, USA. A fast integrated APL has a lot to offer Finance. Currently, most decision support is done with a combination of spreadsheets, relational databases and C routines to glue them together and write analytics. The spreadsheets and RDBMSs are limited in speed and representation — and C is too low-level for high productivity in writing analytics. This paper is about a modern APL, Analytic Platform, which extends the spreadsheet and relational data models and uses C for things that C is good for — namely, intricate system or time-crucial functions. Thus Analytic Platform uses C the way C programs used to use machine code. The list/vector processing language in Analytic Platform is called K. An earlier version of the Analytic Platform called A+ has been in use at Morgan Stanley for 6 years.

APL in Japan Kyozuke Saigusa, Japan. Dr. Saigusa presents a personal survey of the history of APL usage in Japan from 1971 to the present day. His survey, focusing mainly on IBM APLs, reflects the general acceptance of APL in Japan. Dr. Saigusa describes many significant applications, along with some technical problems unique to Japan. After showing the beginning, the growth, the blooming and the decline of APL use the discussion ends with an optimistic vision for the future and a request to better understand the specific requirements of APL for the Japanese market.

The History of APL2 Jim Brown, USA. Dr. Brown wants to remind us of the evolution of APL2 and the many contributions which served to develop the language: Someone said that creativity is the art of remembering what someone has said but forgetting that someone else said it. In this paper I examine the origins of the major ideas that formed APL2 and try to remember who said them. Hundreds of people contributed their thoughts and criticisms. By mentioning some key contributors I don’t mean to slight the others. APL2 was the work of many people and I was the organizer and integrator.”

Closing session: Acquisition and Processing Data from Normal and Cancer Cells Dr. Ivar Giaever, Nobel Prize, Rensselaer Polytechnics Institute, USA. It is common today to control an experimental setup using a dedicated personal computer which is used to both acquire and process the data. Many computer programs exist for this purpose, but APL can easily be adapted to do this task. In this talk, a novel system will be described that can obtain morphological data from cells in tissue culture. When mammalian cells are cultured on small gold electrodes, they will alter the impedance of the electrodes. By measuring the impedance, it is possible to get quantitative data with regards to both the dynamic and static behaviour of the cells. During the experiments, the impedance is plotted on the computer screen, and afterwards the data is analyzed using Fast Fourier Transform, Hurst Rescaled Range Analysis or other mathematical techniques.

Accepted papers¶

A Risk Management Prototype using Object-Oriented APL M. Gfeller. After a quick introduction to Financial Risk Management, I show that Object-Oriented Technology (OOT) fits well with the demands of Risk Management. I then outline the analysis, design, and implementation of a prototype system for Risk Management using both OOT and APL. I describe a Dyalog APL implementation of OOT, and conclude by wondering whether APL is really suitable for OOT.

A Rating System for Qualification Analysis with Application to Personnel Evaluation A. Kononov. An APL-based expert opinion processing system provides a statistical evaluation of information and prepares both an assessment of attributes sufficiency and a textual description or testimonial.

Geometric Data of Cross-sections of Beams H. Besier. This paper describes a program that calculates the geometrical characteristic values of cross-sections of beams with complex shapes (bars or thin-walled profiles with any number of holes in the profiles). After a short introduction describing the calculations involved for moments of area, we will introduce the capabilities of every component of this program and we will make a survey of the whole system via a flow chart. At the end of the paper we will summarize the main features of the program and mention some possible extensions.

Simulation of a Mail Warehouse: An APL2 Solution for a Large Company A. Skomorokhov. The paper describes an APL2 solution to the problem of mail and document distribution in a large (33,000 employees) company. A simulation of a fully-automated warehouse for delivery of business papers is presented and the difficulties of simulation software design are considered.

Time Series Forecasting Using Neural Networks T. Kolarik and G. Rudorfer. Artificial neural networks are suitable for many tasks in pattern recognition and machine learning. In this paper we present an APL system for forecasting univariate time series with artificial neural networks. Unlike conventional techniques for time series analysis, an artificial neural network needs little information about the time series data and can be applied to a broad range of problems. However, the problem of network “tuning” remains: parameters of the back-propagation algorithm as well as the network topology need to be adjusted for optimal performance. For our application, we conducted experiments to find the right parameters for a forecasting network. The artificial neural networks that were found delivered a better forecasting performance than results obtained by the well-known ARIMA technique.

A Parallel Correlation-based Algorithm in J learns Neural Network Connections A. Skurikhin and A. Surkan. This paper describes and demonstrates a J language implementation of a correlation-based algorithm as it is applied to learning feed-forward and recurrent neural networks. Alopex, the learning algorithm, can be used as a tool for solving a wide range of problems for complex systems including: modelling, prediction, diagnosis and pattern recognition. Instead of relying on error gradients, the Alopex algorithm learns on the basis of local correlations between the changes in individual weights and the changes of a global error measure. The algorithm is useful for optimising networks having arbitrary transfer functions distributed over its individual neurones and for minimising a large class of error measures. Alopex has the advantage that it updates the weights in a network simultaneously (i.e. in parallel) requiring only local computations. This permits a complete parallelisation of the algorithm represented in the J language implementation of this demonstration. This J program for the algorithm is applied to the real-world econometric problem of predicting the gas sendout or consumption value one or more days after the end of the time series record.

Knowledge Representation in Expert Systems in a Linguistic Form J. De Meyer. A good expert system needs to be as close as possible to the knowledge and data manipulation of the expert himself. Often experts use linguistic instead of numerical values. When input data are mostly qualitative and are based on subjective knowledge of experts, the Fuzzy Set Theory is a solid mathematical model to represent and manipulate these data. APL arrays, scalars, vectors, matrices, operands, etc. provide powerful means for implementing fuzzy sets. In this paper we will show the capabilities of APL to represent knowledge in a linguistic form.

An Implementation of Prediction Calculation Using APL and Clipper I. Luneva, S. Obraztsov and A. Shimkevich. A complex algorithm for prediction calculation based on non-linear algebraic models has been implemented using Clipper for the user-interface and APL*PLUS for calculations. A currency exchange rate system has been implemented as a DOS application for Windows with dynamic data exchange.

APL and Economic Education A. Buzin. The teaching of various Socio-Economic models formerly viewed as ‘bourgeois theories’ is enhanced by the use of APL in the development of elementary mathematical models including function research, matrix algebra, interactive analysis, and graphic interpretations.

VDE: A Virtual Data Engine for APL J. Merrill. This paper discusses the development history of a data-handling mechanism known as VDE, which stands for Virtual Data Engine. My client, a major international bank, has been able to take advantage of the ease with which a flexible APL-based design can be extended into areas not even imagined during initial planning.

Is APL a Competitive Language in the International Olympiads in Informatics I. Mironov and V. Bargatchev. After participating in the International Olympiads for Informatics the authors the authors evaluated the effectiveness and appropriateness of APL in the solution of the same set of problems.

The Impact of APL on First Year Mathematics W. Spunde. In 1993, the Australian Federal Government’s Committee for the Advancement of University Teaching (CAUT) funded a project to develop an integrated computationally rich first year mathematics course built on the use of APL as an integral part of the instruction. This article outlines how the standard syllabus was amended to respond to the available computing power for elucidating mathematical topics and what mathematical skills were improved or neglected by the novel presentation.

Application of an APL Formula Manipulation System to Different Models of Computer M. Kikkawa, K. Sugawara and Y. Washio. In the course of studying the application of APL to high school education to help students in their understanding of mathematical concepts, many teachers encountered difficulties in that the results of the study could not be used in their classes because the school had provided them with models of computer which cannot run APL. The authors and their co-workers found a way to solve such a difficulty by porting the study file from NiHonGo APL to I-APL. This paper discusses problems encountered in the classrooms, the new system applied to three popular models of personal computers, the actual use of the APL study file in eight schools in learning factorisation, and evaluation of both teachers and students.

Report of the Kronstadt Project B. Shpigel and T. Guryeva. During the APL92 international conference, held in St. Petersburg in July 1992, a conjunction of several independent incidents led to a fruitful and interesting APL educational project at the school of Kronstadt.

Probabilistic Studies of I CHING with the Help of APL P. Luksha. I Ching, the ancient Chinese fortune-telling book, is the basis of the whole Chinese culture. It consists of 64 6-line pictographs called hexagrams, and commentary texts for them. The author presents mathematical problems connected with I Ching and provides their solutions via APL functions.

Useful Extensions of Matrix Divide N. Thomson. This paper show two generalisations of the use of the APL2 domino function. Domino has two roles, namely the solution of non-degenerate systems of linear equations, and the estimation of linear least squares fits. The duality between these is exploited to extend domino, using Newton-Raphson iteration to find roots of non-linear systems of equations on the one hand, and to perform non-linear curve fitting on the other. APL2 operators are very appropriate, and the J approach is also described. Simple idioms give consequent fitted values and Analysis of Variance calculations. A compact tool-kit is given summarising in parallel columns the APL2 and J code given in the paper. The brevity of the latter speaks for itself.

Alberth’s Method for the Parallel Iterative Finding of Polynomial Zeros T. Chi Chen. Alberth’s method for iteratively finding polynomial zeros using simultaneous interacting guesses is seen to be a special case of Newton’s method with parallel anticipatory implicit deflation, changing the single polynomial problem into that for a system of rational functions. It subjects all iterates to the same iterative process, and discourages coincident convergences, at the cost of greater computing labour and slow convergence to multiple zeros. Serial methods interspersed with standard deflation prove much faster though suffer from uneven roundoff error accumulation. Factors affecting convergence include initial guess choices, problem symmetry, and sluggish clusters. The experimental APL program PAIDAX installs the proposed asymmetric initial guesses and immediate updating strategy. Tests show that the favoured scheme does provide welcome asymmetry and speedup, though convergence to multiple zeros remains slow.

Statistical Simulation and Illustration Using the APL2/AGSS Computing Environment P. Lewys. The purpose of this paper is to show that multi-factor statistical simulations can now be very easily and conveniently implemented in the APL2/AGSS computing environment. In particular, this environment has all of the necessary statistical methodology and the colour graphics facilities one needs to delineate the effect of various factors in a statistical simulation. Moreover the main reason for using a batch language like FORTRAN — computing speed — is no longer a deterrent to using an interactive environment like APL/AGSS. With this APL/AGSS version of SIMTBED, most statistical concepts can be illustrated interactively on-line while teaching it a fast microprocessor is available. Several examples are given.

Analysing Time-Series in APL A. Sykes. This paper considers the provision of time-series software for statistics students to use. Specifically it discusses some appropriate graphical output, and the fitting of ARIMA models using a version of Marquardt’s Non-Linear Least Squares algorithm.

Maps of Science Construction using APL and Graphics B. Makeev and A. Zueva. Maps of Science are produced using APL and graphics. These Graphic representations disclose the interlinks between branches of atomic science and the rate of interpenetration between these branches.

The APL Theory of Human Vision G. Langlet. Vision is considered as a mechanism that deals with information processing by a living organism. The conjunction of several hypotheses led to a single theoretical model of quasi-automatic integration of information, which processes compression and orthogonal transforming, without involving any classical calculation; the model that was worked out completely with APL as the main research tool, will be exposed in APL also. It is based on special properties of matrices in modulo-2 integer algebra; the paper covers: considerations on information coding, explanations about the role of patterns formed by retina receptors, about the behaviour of the electrons within the main information-processing molecule (retinal) and about the transmission of information in synapses as an annexe.

APL for Statistical Research of Turbulence Structure during the Dust Storm A. Mukanova. The Aral dust storms are statistically analyzed. Special attention is given to experimental and theoretical investigations of desert aerosol transport problems for quantitative description of dust transfer from source to precipitation areas by wind.

LiSEB: A Language for Modelling Living Systems with APL2 P. Bottoni, M. Mariotto and P. Mussio. The paper discusses the design and implementation of LiSEB, a class-based language built on top of APL2 to respond to challenges posed by modelling living systems from a medical point of view. LiSEB capitalises on several features of APL and on some lessons learned from its history. Living systems are modelled as open systems in which concurrent mobile autonomous agents interact. Modelling of these properties required extensions of traditional object-oriented paradigms and of their previous translations under APL: a) every object performs a sequence of actions dynamically modified to adapt to circumstances; b) a new policy of message management is introduced uniformly encompassing broadcast and directly addressed communication. An example of a simulation using LiSEB illustrates these concepts.

A New Graphical Interface for APL Dick Bowman. Historically, APL has had two significant advantages over many other problem-solving computational tools. The first of these is that APL’s language elements have a consistency and breadth that remains unchallenged, at least for the tasks for which those elements were designed. It may be arguable that some widening of scope beyond these linear algebra origins is long overdue, but that is beyond the scope of this paper. The other historic advantage of APL, its provision of an interactive problem-solving environment, has long since been eroded. The purpose of this paper is to explore some possible ways in which the face that APL presents to its users could usefully be enhanced. These ideas are presented from the viewpoint of an APL user and there is no consideration of the practicality of their implementation into viable APL interpreters.

A New Development in the Technique of APL Modelling using APL*PLUS and a C++ Compiler D. Gusev and I. Pospelov. A compiler that can convert non-interactive APL*PLUS programs into machine-executable code uses C++ as an intermediate language. The compiled code is machine-independent and runs three times faster than the original APL program.

An Object-Oriented APL D. Selby. Object-oriented programming has become an important, and accepted part of the computer software industry. Nearly every new operating system that has recently arrived or is scheduled to arrive soon is object-oriented in nature. If APL2 is to continue to be a mainstream computational environment, it is very important that it becomes object-oriented. Finding a way that APL2 may sit comfortably in these new operating systems is important. Unfortunately, in both implementations and current programming teaching philosophy, existing APL2 does not easily offer the end-user the ability to adopt this new approach. This paper proposes some small changes to the basic APL2 language definition which will permit objects to be defined and used. It also suggests some major changes to the end-user’s programming methodology.

Restriction Riddance: More Power to Parentheses F.H.D. van Batenburg. This paper proposes to enhance the power of parentheses. Instead of restricting parentheses to expressions that yield an argument of a function, as in “(2-3)+6”, I propose to use parentheses also for arguments of operators. If operators are allowed to have not only functions, but also parenthesised expressions as arguments, APL expressions will be more powerful. It will be shown that this proposal enables the programmer to make various types of expressions more readable and even make it possible to experiment with different types of structured programming constructs.

Programming in J/Windows C. Burke. J has been available as shareware for the last four years. The core language has been largely complete from the outset, but it is only comparatively recently that the implementation (at least for the PC) has included features needed for serious application development. The most significant change was the introduction of Windows support in release 6, which for the first time allowed full screen applications to be written in J. Following the release of 6.2, which included many additions to the Windows support and a much better session manager, I started using the Windows version of J in small applications, for which I -would normally have used APL. Since then I have switched over to using J for almost all development, except for client work where APL is specifically required. I believe that J is largely comparable to commercial APLs for application development and will soon be superior. In this paper, I will describe programming the J/Windows interface, and discuss some practical considerations in using it. The focus throughout will be on Windows — in what follows, for ‘J’, read: ‘J/Windows’.

TAGS: Trains, Agendas and Gerunds K. Iverson and R. Hui. Trains, agendas , and gerunds have been available in J for some time. This paper reviews their definitions and uses, and presents some new extensions that enhance their utility.

Posters¶

APL and Demographic Models Umberto Piasentin, Maria Stovall, David Lanari. A demographic projection model for regional planning will be presented. The base year population of the region (about one million inhabitants) was disaggregated into age/sex categories (one hundred annual age classes by each sex). The APL language was chosen for its strength to develop the model and to realise a friendly interface for the planner. The model can be used not only for a neutral projection of the population growth but also for a simulation of important changes in the demographic system (migration, fertility,...).

APL as Multi-lingual Communication Tool: A Small Machine Translation System for English-Esperanto and Japanese-Chinese Toshio Nishikawa. A small machine translation system was developed in APL. It demonstrates that APL is capable of manipulating character string data as well as numerical data, applied even to machine translation; and simultaneously it attracts APLers’ interests to APL as an international communication tool. For the time being, as a prototype of machine translation, we installed the easiest language pairs, English-Esperanto and Japanese-Chinese translations.

Building Science Models in APLomb Ian A Clark. APLomb is an educational software construction set for the Apple Macintosh. It is not just one product but a range of products based on interconnecting parts. The parts, called “kompos” (derived from “component”), provide ready-made screen furniture such as gauges, arrays of icons, coloured pictures, sounds and sprites. A powerful function-definition language (ISO-standard APL) makes the kompos do useful things. You can liken APLomb to a toy construction set based on snap-together parts.