## Seminars

Seminar
TITLEMinimization of a Regression Function with Application to Problems in Machine Learning
SPEAKER Prof. Sid Yakowitz
Systems and Industrial Engineering Department, University of Arizona, Tucson, Arizona, USA
WHEN2:15 PM, Monday, January 20, 1997
WHERETheatre B, Richard Berry Building, The University of Melbourne
ABSTRACTClassical tasks in machine learning as well as many popular parlor games and operations research problems can be abstracted to the statistical task of finding the minimum of a discontinuous regression function on the basis of noisy measurements. In applications, the regression function is erratic, and popular techniques such as stochastic approximation and simulated annealing are inappropriate.

Ideas and convergence properties from the realm of random search are surveyed, and results of automatic learning experiments with the 8-puzzle, blackjack, and epidemic control are shared. Through case studies and theory, the speaker will contend that this problem area has deep ramifications with respect to bringing statistics and computers to bear on practical activities. Moreover, stochastic minimization raises tantalizing theoretical questions.

CONTACTMoshe Sniedovich, Tel: 9344 5559, E-mail: moshe@maths.mu.oz.au
We regret that we cannot offer free parking to participants.

Seminar
TITLE On smoothing continuation methods for variational inequalities and nonlinear complementarity problems.
SPEAKER Dr Xiaojun Chen
School of Mathematics, University of New South Wales
WHEN 3.15PM, Thursday, 3 July 1997
WHEREThomas Cherry Room, Richard Berry Building, University of Melbourne
ABSTRACTThis talk presents an overview of the recent developments in smoothingmethods for the variational inequality problem (VIP) and the nonlinear complementarity problem (NCP). The VIP and the NCP with a continuous mapping f from R^n into itselfcan be written as the system of equations F(x)=0. Here F:R^n\to R^nis locally Lipschitzan continuous but possibly not differentiable. Smoothing methods use parametric differentiable functions to approximate F and trace a trajectoryconsisting of solutions of the smooth systems until the parameter attains 0.We will discuss the following features of the smoothing methods for the NCP:
1. It is well defined and any accumulation point is a solution x* of the NCP if f is a P_0 function.
2. The sequence {xk} for k=1,2 ... generated by the method is bounded and globally converges to x* if f is a P_0 and R_0 function or f is a monotone function with a feasible interior point.
3. The sequence {xk} is locally superlinearly convergent to x* if the generalized Jacobian of F at x* is nonsingular.
4. It finds a solution in finite steps if f is an affine function.
5. It solves a system of linear equations at each step.
6. It starts from any point in R^n..
CONTACTDr Danny Ralph, E-mail: moshe@maths.mu.oz.au

Seminar
TITLE Bootstrap tests for common structure in a family of (auto) regressive models
SPEAKER Professor Qiwei Yao
University of Canterbury
WHEN 23 July 1997, 1.00 - 2.30 PM
WHERE RMIT, Department of Statistics and Operations Research 360 Swanston Street, Building 8, Level 9 Room 66
ABSTRACTMotivated by some practical problems, two bootstrap tests areproposed for testing the common structure in a family of time seriesmodels. By using the ordinary bootstrap and wild bootstrap alternatively,we are able to test whether the autoregression functions as well as thevariances of error terms are the same. The asymptotic theory has beenestablished which justifies the use of bootstrap methods. The methodsare ready to be applied to general regression setup.

CONTACT Kaye E. Marion, Email: k.marion@rmit.edu.au

Seminar
TITLE Bootstrap tests for common structure in a family of (auto) regressive models
SPEAKER Professor Qiwei Yao
University of Canterbury
WHEN 23 July 1997, 1.00 - 2.30 PM
WHERE RMIT, Department of Statistics and Operations Research 360 Swanston Street, Building 8, Level 9 Room 66
ABSTRACTMotivated by some practical problems, two bootstrap tests areproposed for testing the common structure in a family of time seriesmodels. By using the ordinary bootstrap and wild bootstrap alternatively,we are able to test whether the autoregression functions as well as thevariances of error terms are the same. The asymptotic theory has beenestablished which justifies the use of bootstrap methods. The methodsare ready to be applied to general regression setup.

CONTACT Kaye E. Marion, Email: k.marion@rmit.edu.au

Seminar
TITLE Mathematical Programming Approaches to Machine Scheduling Problems: an Assessment"
SPEAKER Prof. JATINDER N. D. GUPTA
Department of Management, Ball State University, Muncie, IN 47306, USA
WHEN Thursday 24 July 1997, 12.00 - 13.00
WHERE Classroom D, Richard Berry Building, University of Melbourne
ABSTRACTThis paper considers the combinatorial optimization problemsassociated with scheduling a given number of jobs on a specifiednumber of machines to optimize a some well defined measure ofperformance. These problems can be formulated as mixed-integerprogramming problems. Starting with a brief overview of themachine scheduling problems, some recent developments in the useof mathematical programming formulations for the flowshopproblems are explored and an assessment is made as to thecomputational feasibility of using such approaches for solvingpractical machine scheduling problems.

CONTACTDr Natahsia Boland, Department of Mathematics and Statistics, University of Melbourne

Seminar
TITLE EXCURSIONS IN SCHEDULING THEORY
SPEAKER Prof. JATINDER N. D. GUPTA
Department of Management, Ball State University, Muncie, IN 47306, USA
WHEN Friday 25 July 1997, 11.30 - 12.30
WHERE Swinburne University of Technology, Engineering Building, Room EN309
ABSTRACTThis seminar is designed to provide an excursion into various scheduling problems arising in the manufacturing environment and possible approaches that can be taken to solve some of these problemsThe seminar reviews the literature from one specific perspective from one researcher`s viewpoint and describes the mental and professional development of one person in understanding the complex field of production management in general and scheduling theory in particular.This seminar will cover the fundamental frameworks of scheduling theory, outlining various approaches that can be taken to solve (optimally or approximately) such problems, and the difficulties arising in their practical use. Specific detailed techniques of solution algorithms will not be covered in this seminar as they are better discussed in individual encounter.

CONTACT David Noble (9214 8266 or dnoble@swin.edu.au)

Seminar
TITLE SIMULATING RUIN PROBABILITIES AND THEIR DERIVATIVES
WHEN5:30PM, August 1, 1997
WHEREClassroom D, Richard Berry Building, The University of Melbourne
ABSTRACT We discuss the problem of estimating ruin probabilities by simulation, as well as their sensitivities with respect to the claim rates. When ruin is not certain, directsimulation of the system is not applicable, since there is no terminating criterion. Three alternative methods are discussed: Importance Sampling for rare events, a dual storage process, and direct simulation of the convolutionformula.

RPA can be used to obtain pathwise derivatives w.r.t. the claim rate for the first two methods, and we use a likelihoodratio approach for the third method. We shall finish withnumerical comparisons of the efficiency of the estimators.

CONTACTMoshe Sniedovich, Tel: 9344 5559, E-mail: moshe@maths.mu.oz.au
We regret that we cannot offer free parking to participants.

Seminar
TITLE Costing, Pricing and Inventory Control for Dura-Wills
SPEAKER Peter Brooksbank, Managing Director, Euratech LtdDudley Foster, DNF Decision Sciences
WHEN5:30 PM, Wednesday, August 20, 1997
WHERE RMIT Room 8.9 66 (Building 8, Level 9, Room. 66)

Seminar
TITLE Is it a hard problem, really?
(Noddy and Big Ears' guide to abstract complexity theory)
SPEAKERDr Liz Sonenberg
Department of Computer Science, University of Melbourne
WHEN5:30 PM, Thursday, September 11, 1997
WHEREClassroom D, Richard Berry Building, The University of Melbourne
ABSTRACTIn this presentation I will introduce basic ideas of complexity theory dating back to the 70's (i.e. ways of classifying problems into degrees of difficulty) and also discuss some recent advances in this area (i.e. updating with a somewhat idiosyncratic view of interesting results from the 90's).For the bulk of the presentation, no particular prior knowledge of this area of computer science will be assumed.

CONTACTMoshe Sniedovich, Tel: 9344 5559, E-mail: moshe@maths.mu.oz.au
We regret that we cannot offer free parking to participants.

Seminar
TITLE SIMULATED ANNEALING: SEARCHING FOR AN OPTIMAL TEMPERATURE SCHEDULE
SPEAKERDr Harry Cohn and Mark Fielding
Department of Mathematics and Statistics, University of Melbourne
WHEN5:30 PM, Thursday, September 18, 1997
WHEREClassroom D, Richard Berry Building, The University of Melbourne
ABSTRACTMost of the theoretical literature on simulated annealing deals with aproperty called convergence which asserts that the simulated annealingchain is in the set of global minima states of the objective functionwith probability tending to one. However, in practice the convergentalgorithms are considered too slow whereas a number of non convergent onesare usually preferred. We attempt a detailed analysis of varioustemperature schedules.Examples will be given when it is both practically and theoretically justifiedto use boiling, fixed temperature and even fast temperature schedules whichhave a small probability of reaching global minima. Applications to travellingsalesman problems of various sizes are also given.
CONTACTMoshe Sniedovich, Tel: 9344 5559, E-mail: moshe@maths.mu.oz.au
We regret that we cannot offer free parking to participants.

Seminar
TITLE Some Applications of MIP and Heuristics to optimisation
SPEAKERSBertil Marksjo and Robert Parker
CSIRO - Division of Building Construction and Engineering
WHEN5:30 PM, Wednesday, October 15, 1997
WHERE RMIT Room 8.9.66
ABSTRACTApplications of MIP and Heuristics on the optimisation of the use ofinfrastructure for telecommunication networks, and planning for watersupply infrastructure and operational policies. This talk will highlight the merits of contrasting and combining complementary techniques.

AGM
6:30PM, Wed 19 March 1997
Followed by a talk by Dr Bruce Craven. Make sure you put this in your diary to hear Bruce give an anecdodatal account of the formation and history of the Melbourne Chapter of ASOR.
Nomination forms will be circulated nearer to the AGM.

Student Conference
Wed 17 September, 1:30-5:30
Swinburne University of Technology
TAFE Building D, Room D121

### AIOR WORKSHOPS

Two further Workshops in this series will be held at Swinburne University of Technology (Hawthorn Campus):

Wed October 1: Financial Modelling (Presenter: Dudley Foster)
Thurs October 2: Forecasting (Presenter: Harry Gielewski)

Cost for 1 day will be $325 for ASOR Members ($350 for non-members)
Cost for 2 days will be $600 for ASOR Members ($650 for non-members)

For further information and course leaflets, contact:

David Noble
Tel: 9214 8266
Fax: 9819 0821
Email: dnoble@swin.edu.au

### EXECUTIVE COMMITTEE 1997 - 1998

Chairperson:

David Noble
School of Mathematical Sciences
Swinburne University of Technology
P O Box 218
HAWTHORN VIC 3122
Australia
E-Mail: dnoble@swin.edu.au
Tel: (W) + 613 9214 8266
FAX: 9819 0821

Vice-Chairperson:

Harry Gielewski
28 Kennedy Street
RESERVOIR VIC 3073
Australia
E-mail: harryg@matilda.vut.edu.au
Tel: (W) + 613 9350 4726
Mobile: 0414 650 110

Secretary:

Baikunth Nath (Dr)
School of Computing and Inform. Technology
Monash University
CHURCHILL VIC 3842
Australia
E-Mail: B.Nath@fcit.monash.edu.au
Tel: (W) + 613 9902 6468
FAX: 9902 6842

Treasurer:

Lutfar Khan (Dr)
Department of Computer and Mathematical Sciences
Victoria University of Technology
P O Box 14428, MCMC
MELBOURNE VIC 8001
Australia
E-Mail: Khan@matilda.vut.edu.au
Tel: (W) + 613 9688 4687
FAX: + 613 9688 4050

Committee:

Peter Cerone (Assoc Prof)
Department of Computer and Mathematical Sciences
Victoria University of Technology
PO Box 14428 MCMC
MELBOURNE VIC 8001
E-Mail: pc@matilda.vut.edu.au
tel: (W) + 613 9688 4689
FAX: + 613 9688 4050

Paul Lochert (Assoc Prof)
Department of Mathematics
Monash University
P.O. Box 197
CAULFIELD EAST VIC 3145
Australia
E-Mail: P.Lochert@sci.monash.edu.au
Tel: (W) + 613 9903 2647
FAX: + 613 9903 2227

Kaye E. Marion (Ms)
Department of Statistics & OR
RMIT
360 Swanston Street
MELBOURNE VIC 3000
Australia
E-Mail:
Tel: (W) + 613 9660 3162
FAX: + 613 9660 2454

Moshe Sniedovich (Dr)
Department of Mathematics and Statistics
University of Melbourne
PARKVILLE VIC 3052
E-Mail: moshe@mundoe.maths.mu.oz.au
moshe@mundoe.maths.mu.oz.au
WWW: http://www.maths.mu.oz.au/~moshe/
Tel: (W) + 613 9344 5559
FAX: + 613 9344 4599

Patrick Tobin
School of Mathematical Sciences
Swinburne University of Technology
P O Box 218
HAWTHORN VIC 3121
Australia
E-Mail: ptobin@swin.edu.au
Tel: (W) + 613 9214 8013
FAX: + 613 9819 0821

Dudley Foster
23 Wolseley Crescent
BLACKBURN VIC 3130
Australia
Tel: (W) + 613 9248 1069Mobile: 0417 342 272

Harry Gielewski
28 Kennedy Street
RESERVOIR VIC 3073
Australia
E-mail: harryg@matilda.vut.edu.au
Tel: (W) + 613 9350 4726
Mobile: 0414 650 110

Student representative:

Luminita Baloi
Level 17, 452 Flinders Street
MELBOURNE VIC 3000
Australia
E-mail:
Tel: (W) 9229 6271
FAX: 9229 6292

Ex-Officio:

Stephen Weal
Division of Swinburne at Lilydale
Swinburne University of Technology
Locked Bag 218
Lilydale VIC 3190
Australia
E-Mail: sweal@swin.edu.au
Tel: (W) + 613 9215 7131
FAX: + 613 9215 7070

Co-opted:

Santosh Kumar (Professor)
Department of Applied Mathematics
National University of Science & Technology
P O Box 346, Bulawayo
ZIMBABWE
E-Mail: MANGENA@esanet.zw

### Consulting Services

• OR Solutions: Problem Solving through Quantitative Analysis
P.O.Box 2086