close

Вход

Забыли?

вход по аккаунту

?

Book Review Neural Networks for Chemists. An Introduction. By J. Zupan and J. Gasteiger

код для вставкиСкачать
BOOKS
on the other. that their readers should
master quantum mechanics up to the
(spin) density matrix, the Hilbert space
and the like. In my opinion it is a complete
illusion to hope that someone who needs
an introduction to complex numbers is
able to appreciate any exposition of quantum mechanics. I am afraid, therefore,
that Hennel’s and Klinowski’s book is of
little help to those who have not previously followed a course in quantum mechanics. For those who have done so (and have
worked through their homework problems) the book may be a useful revision.
and they may like to see how quantum
mechanics is applied cwx-tly and simply
and at great length to practical cases.
However, a major problem is that the
book does not lead the reader far enough.
Almost all the fundamentals remain on an
educational, a cultural level. They d o not
open the door to practical modern N M R .
On the whole the book breathes the air of
the fifties. maybe sixties. Pake’s ingenious
but hopelessly obsolete “high frequency
bridge“ is described to illustrate how
N M R is detected. Pulse and F T N M R are
discussed but no modern probehead or related instrumentation is shown. Of the total of 54 references no less than 37 appeared in o r before 1967. and 19 of these
are from the fifties. All those from after
1980 are monographs and textbooks.
Abragani appears under 1983, whereas I
bought my copy of his book in 1961.
The final chapter is on “Nuclear Magnetic Relaxation”. It brought me an almost nostalgic re-encounter with Gierer
and Wirtz (1953) and their microviscosity
coefficients. I had hoped for a more up-todate climax. I recommend the book for
introductory N M R teaching. either in a
class or in a laboratory course, and to all
those who d o practical N M R work and
desperately feel that they d o not really understand the fundamentals of their experimental tool. But they should not expect a
link between these fundamentals and their
own work.
U. Hueherlen
Max-Planck-Institut
fur medizinische Forschung
Heidelberg (FRG)
Neural Networks for Chemists. An
Introduction. By J. Zupan and J.
Gasteiger. VCH Verlagsgesellschaft, Weinheim/VCH Publishers,
New York, 1993. 305 pp., hardcover
DM 138.00, paperback DM 68.00.
-ISBN
3-527-28592-X/1-56081791-7 (hardcover), 3-527-28603-91
1-56081-793-3 (paperback)
To explain precisely the basic concepts
of a new interdisciplinary area of research
and communicate them to a wide readership is a difficult task. Jure Zapan and
Johann Gasteiger have managed to d o
this very successfully. This book offers a
sound introduction to artificial neural
networks, with insights into their architecture, functioning, and applications, which
is intended not only for chemists but also
for scientists from other disciplines. The
capabilities and limitations of different
systems are compared and evaluated. The
authors have sensibly limited their treatment to a few carefully chosen and clearly
distinguished types of networks, explaining
how they are used for classifying, modeling, and analyzing molecules and their
properties, for evaluating visual images
(e.g.. to interpret spectra o r to analyze
structure-activity relationships), and for
controlling chemical reactions. The book
is written in a readable style and is well
structured. The reader’s task is made easier by providing a short summary of the
main learning objectives at the start of
each chapter, and by the use of easily remembered mnemonics. The treatment is
divided into two parts, one concerned
with theory and the other with describing
different types of applications. This arrangement means that, after studying the
clear and very informative introduction to
the topic, the reader can look at the examples of applications and assess the suitability of the chosen method in each case,
possibly even suggesting an alternative solution. The chapters on Kohonen networks and counterpropagation systems
are especially good. The examples of applications, such as the representation of
the three-dimensional electrostatic potential of a molecule in the form of a two-dimensional map using the Kohonen approach, surprise one at first by the
simplicity of the concepts, but this is in
itself an illustration of one of the strengths
of artificial neural networks. The authors
explain clearly and concisely how these
principles can be applied in order to make
an appropriate selection of representative
data for an analysis using neural networks. The reader is not distracted from
the core issues by too many literature references in the text; instead the authors
give a useful selection at the end of each
chapter.
In OUJ view there are only a few aspects
of neural networks that have not been fully covered in the book. The authors have
not discussed effective algorithms for the
systematic optimization of network architectures --for example, genetic algorithms.
Also, in restricting the discussion to the
back-propagation algorithm for multipositional networks the authors have covered only a part (admittedly by far the
largest) of the range of reliable strategies
for supervised learning. A similar criticism
applies to the choice of applications. In
particular. the analysis of biological
macromolecules receives only marginal
treatment, being limited to predicting the
three-dimensional structures of proteins
from their amino acid sequences. Here it
would have been desirable to include
some more examples -e.g., the use ofneural networks to analyze data banks- -as
these methods have already advanced
considerably and much effort is being put
into further developments. However. additional topics such as these can no doubt be
included in a second edition. This is a very
well-written guide for anyone who wants
to make greater use of information theories in chemistry. The excellent quality of
the contents and the presentation should
ensure that it reaches a wide international
readership.
Gisbert Schneidu. Paul Wrerle
Institut fur Medizinischel
Technische Physik und Lasermedizin
der Freien Universitiit Berlin (FRG)
Homogeneous Transition Metal Catalyzed Reactions. Edited by W R.
Moser and D.W Slocum. (Series:
Advances in Chemistry, Vol. 230.)
American Chemical Society, Washington, DC, 1992. 625 pp., hardcover $! 139.95.-ISBN 0-8412-2007-7
This book contains a substantial number of the presentations from the symposium on “New Science in Homogeneous
Transition Metal Catalyzed Reactions”,
held during the 199th National Meeting of
the American Chemical Society, in Boston,
Massachusetts, April 22-27, 1990.
About one third of the chapters have
been written by chemists working in industrial laboratories. The authors are in
general very well known for their contributions to the field. The fact that the papers are subjected to the editorial standards of the ACS is a further warranty for
the high quality of the book.
Документ
Категория
Без категории
Просмотров
0
Размер файла
150 Кб
Теги
introduction, chemists, book, neural, network, zupan, gasteiger, review
1/--страниц
Пожаловаться на содержимое документа