Abstract / Description of output
The
analysis
of
charge exchange recombination
spectra
represents
a
very
&allengin%
problem
due
to
the presence
of
many overlapping spectral lines. Conventional approaches
are
based
on
iterative least-squares optimization and suffer from the
two
difficulties
of
low
speed
and
the
need
for
a
good
initial
approximation
to
the
solution.
This
[suer
problem
necessitates
considerable human supervision
of
the
analysis
procedure.
In
this paper
we
show
how
neural
network techniques allow charge exchange
data to
be
analysed very rapidly,
to
give an approximate solution without the
need
for
supervision
The
network approach
is
well
suited
to
the
fast
intershot analysis
of
large volumes
of
data,
and
can
readily
be
implemented
in
dedicated hardware
for
real-time applications.
The
neural network
can
aIso
be
used
to
provide
the
initial
guess
for
the standard least-squares algorithm
when
high accuracy
is
required.
| Original language | English |
|---|---|
| Title of host publication | Proceedings 14th European Conference on Controlled Fusion and Plasma Physics, Madrid |
| Pages | 765-773 |
| Number of pages | 9 |
| Publication status | Published - 1 Jan 1987 |