<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
<title>South African Computer Journal 1991(5)</title>
<link href="https://ir.unisa.ac.za/handle/10500/23878" rel="alternate"/>
<subtitle/>
<id>https://ir.unisa.ac.za/handle/10500/23878</id>
<updated>2026-05-05T15:30:42Z</updated>
<dc:date>2026-05-05T15:30:42Z</dc:date>
<entry>
<title>The physical correlates of local minima</title>
<link href="https://ir.unisa.ac.za/handle/10500/23955" rel="alternate"/>
<author>
<name>Wessels, LFA</name>
</author>
<author>
<name>Barnard, E</name>
</author>
<author>
<name>Van Rooyen, E</name>
</author>
<id>https://ir.unisa.ac.za/handle/10500/23955</id>
<updated>2018-05-20T01:00:42Z</updated>
<published>1991-01-01T00:00:00Z</published>
<summary type="text">The physical correlates of local minima
Wessels, LFA; Barnard, E; Van Rooyen, E
The training  of neural-net classifiers is often hampered by the occurrence  of  local minima, which results in the attainment of inferior classification performance. We study the problem of local minima in order to devise means of alleviating it. In order to establish a better understanding of the problem, the nature of the physical states of neural nets stuck in local minima, is investigated. We show that the occurrence of a local minimum in the crite­rion function can often be related to specific patterns of defects in the classifier. In particular, three main causes for local minima are identified. Such an understanding of the physical correlates of local minima is important, since it suggests sensible ways of choosing the weights from which the training process is initiated.
</summary>
<dc:date>1991-01-01T00:00:00Z</dc:date>
</entry>
<entry>
<title>Knowledge representation using formal grammars</title>
<link href="https://ir.unisa.ac.za/handle/10500/23954" rel="alternate"/>
<author>
<name>Von Solms, S.H.</name>
</author>
<author>
<name>Ehlers, E.M.</name>
</author>
<author>
<name>Enslin, DJ</name>
</author>
<id>https://ir.unisa.ac.za/handle/10500/23954</id>
<updated>2018-06-06T11:42:53Z</updated>
<published>1991-01-01T00:00:00Z</published>
<summary type="text">Knowledge representation using formal grammars
Von Solms, S.H.; Ehlers, E.M.; Enslin, DJ
In this paper, formal grammars, specifically random context grammars, are used as a vehicle for knowledge representation in expert systems. Random context grammars and the manner in which they are used to represent knowledge are discussed. An example of a knowledge base is given.
</summary>
<dc:date>1991-01-01T00:00:00Z</dc:date>
</entry>
<entry>
<title>An efficient primal simplex implementation for the continuous 2-matching problem</title>
<link href="https://ir.unisa.ac.za/handle/10500/23953" rel="alternate"/>
<author>
<name>Smith, THC</name>
</author>
<author>
<name>Meyer, TWS</name>
</author>
<author>
<name>Leenen, L</name>
</author>
<id>https://ir.unisa.ac.za/handle/10500/23953</id>
<updated>2018-05-20T01:00:46Z</updated>
<published>1991-01-01T00:00:00Z</published>
<summary type="text">An efficient primal simplex implementation for the continuous 2-matching problem
Smith, THC; Meyer, TWS; Leenen, L
The  continuous  2-matching  problem   {RMP2)  is the  relaxation  of   the  symmetric  travelling  salesman  problem {STSP) used by Padberg &amp; Rinaldi to develop a highly successful branch-and-cut algorithm for the STSP. They used a standard linear program solver for solving RM P2. We note that RM P2 is a generalized network problem with additional special structure and exploit this to provide an efficient implementation of the primal simplex algorithm for RM P2. Our computational experience with the implementation demonstrates that  it is several  or­ders of magnitude faster  than a standard  linear program  solver, suggesting that it should  be worthwhile using this&#13;
implementation in the Padberg-Rinaldi algorithm.
</summary>
<dc:date>1991-01-01T00:00:00Z</dc:date>
</entry>
<entry>
<title>An update on UNINET-ZA: The Southern African Academic and Research Network</title>
<link href="https://ir.unisa.ac.za/handle/10500/23952" rel="alternate"/>
<author>
<name>Shaw, V</name>
</author>
<id>https://ir.unisa.ac.za/handle/10500/23952</id>
<updated>2018-05-20T01:00:45Z</updated>
<published>1991-01-01T00:00:00Z</published>
<summary type="text">An update on UNINET-ZA: The Southern African Academic and Research Network
Shaw, V
The  UNINET  Academic  and  Research  Network  is  a collaborative project among Universities, Research Councils and the FRD for the development, implementation and promotion of an academic and research network of computers in southern Africa, where  it  is  required  as  an  essential  element  of  the region's research infrastructure.
</summary>
<dc:date>1991-01-01T00:00:00Z</dc:date>
</entry>
</feed>
