The next meeting of the Fortran Specialist Group will

be held on Thursday 26 October 1989 from 10.30am to 4.30pm

at BCS HQ, 13 Mansfield Street, London.

        The morning meeting will be normal business including

reports from X3J3 and ISO representatives and also a report

on the Fortran Forum 89 held in London and Edinburgh.

        The subject of the afternoon session is Expert Systems.

Our first speaker is Professor John Nelder of Imperial

College who will introduce us to "Expert Systems: Some gen-

eral aspects, and special properties of statistical front

ends". Our second speaker is Dr. John Chelsom of NAG who

will describe "The Application of Knowledge-Based Systems to

Enhance Existing Fortran Software".

        Please bring this notice to the attention of any col-

leagues likely to be interested: we particularly welcome

newcomers to the Group whether members of BCS or not.

Click here for the minutes of this meeting. Links below are to specific
sections of the agenda and minutes.


10.30 am   1.    Apologies for absence

           2.    Minutes of the last meeting

           3.    Matters arising

           4.    Reports from X3J3 and ISO Representatives

           5.    Fortran Forum 89

                 (a) London 27 September 1989

                 (b) Edinburgh 28 September 1989

           6.    BCS and any other business

2.00 pm          "Expert Systems"

                  Professor J A Nelder - Imperial College

                  Dr J Chelsom - Numerical Algorithms Group

John Young, Secretary.

Tel: 0734 814111 Ext 6292.

Expert Systems:        Some general aspects, and special properties of

                        statistical front ends.

                        by J.A. Nelder, Imperial college

Expert systems involve the coding of expertise as a computer program.

Decisions are needed on (i) how to collect the data, (ii) how to

structure the rules, (iii) what language to code the rules in, (iv) how

to design the user interface, and (v) how to incorporate other relevant

pieces of software.  Experience in developing GLIMPSE will be drawn on,

with a discussion of the special properties of front ends for statistics.

     The Application of Knowledge-Based Systems to Enhance

Existing Fortran Software

by J. Chelsom, The Numerical Algorithms Group Ltd


There are (at least) two ways in which one could talk about the Fortran applications of

knowledge-based systems.  One of the first knowledge-based systems, called CASTNET,

was developed in Fortran during the early l970s at Rutgers University, for the diagnosis of

the eye disease Glaucoma. From this project emerged the knowledge-based system shell

EXPERT, one of several forerunners to the wide range of commercially developed shells

available today. Although most of these shells are now developed in C or as an extension to

Lisp or Prolog environments, there is at least one that is written in Fortran.

The other link between Fortran and knowledge-based systems lies in the use of knowledge-

based systems as front ends to existing Fortran software. Fortran has existed as a

programming language for some 30 years and over that time a vast number of applications

has been written. With the increasing complexity of many fields of science and technology,

it is becoming unreasonable to expect domain experts also to be expert Fortran

programmers, capable of exploiting the accumulated collection of Fortran software. It is in

this context that knowledge-based systems are emerging; as an attractive way of 'front-

ending' existing Fortran applications to make them accessible to a wider audience.

In the field of medicine, there are several projects underway that aim to link knowledge-

based systems with complex physiological models written in Fortran. While the knowledge-

based front manages the user interface and overall consultation strategy, the physiological

model offers advice about the long-term prognosis of the patient and the effects of different

therapy plans. At The Numerical Algorithms Group we are currently developing a

knowledge-based front end to the NAG Fortran Library to assist users with the selection and

use of routines suitable for particular problems. This work is being carried out as part of an

ESPRIT project which has the wider aim of developing a toolkit to assist in the construction

of knowledge-based front ends for any suitable software package.





The 1989 Fortran Forum took place at two venues on two

consecutive days. Wednesday, 27th September in the Fleming

Lecture Theatre in the Engineering Department of University

College, London saw the gathering of 37 Fortran "buffs" to

review the second draft of the Fortran 8x standard. The

following day, Thursday, 28th September in Lecture Theatre B

in the James Clerk Maxwell Building of Edinburgh University

was the turn of the Scottish "section" to review the draft.

Regrettably, only l4 people were able to attend, 5 of whom

had attended the London Forum the previous day. This year's

Forum was organised by the Group itself in conjunction with

BCS at very short notice as the second draft was already two

months into its public review period. The Group apologises

for the lack of publicity for this event but hopes that all

those interested in Fortran and who were unable to attend

the Forum will still find the time to submit their comments.

The draft standard Fortran document can still be obtained

from Karen McPhilemy at BCS HQ for the price of £15.00

each. The closing date for comments on the draft standard

for X3J3 is 24th November 1989, for BSI is 15th November

1989 and for ISO is 1st February 1990. Details of how to

submit comments are given in the front of the document.

1989-1990 PROGRAMME

The official date for the closure of BCS HQ is 30th June

1990 so all our forthcoming meetings will take place at

13 Mansfield Street. The Secretary, John Young, was to have

given a brief talk on the use of Fortran in Seismology at

the AGM in Oxford but it was cancelled due to lack of time.

He suggests that the afternoon session of the June meeting

is a series of short talks given by members of the Group.

If you have a particular interest with a Fortran connection

and would be willing to tell the Group about it, please

contact the Secretary.

The dates of the meetings for the next year are:-

      Thursday, 26th October, 1989     -    Expert Systems

      Thursday, 25th January, 1990     -    Parallel Processing

      Thursday, 5th April, 1990        -    AGM and 20th

                                            Birthday Celebrations

      Thursday, 28th June, 1990        -    Members' Talks