Title: Self-scaling conjugate gradient methods for large-scale optimization
| Speaker: | Mehiddin Al-Baali |
| Affiliation: | Sultan Qaboos University |
| Room: | MC 6486 |
Abstract:
The
conjugate
gradient
class
of
methods
with
inexact
line
search
will
be
analyzed.
To
rectify
certain
drawbacks
of
these
methods,
a
new
self-scaling
strategy
to
the
conjugate
gradient
parameter
will
be
considered.
Hence,
the
practical
descent
property
and
global
convergence
result
of
Al-Baali,
for
the
Fletcher-Reeves
method
with
inexact
line
search,
will
also
be
extended
to
other
conjugate
gradient
methods.
Numerical
results
will
be
described
to
illustrate
the
behavior
of
several
members
of
the
class
of
methods
(in
particular,
those
of
the
well-known
Fletcher-Reeves,
Polak-Ribi?re-Polyak
and
Hestenes-Stiefel).
It
will
be
shown
that
the
proposed
self-scaling
strategy
improves
the
performance
of
conjugate
gradient
methods
in
several
cases.