Skip to content

iteration history

4 messages · Robert, Gabor Grothendieck, Marco Geraci +1 more

#
Dear R Users
I would like to use optim function to optimize a function. I read help but I
couldn't find what I need: is it possible to get information after each
iteration, for example as there is in MATLAB:
                                                        Gradient's 
 Iteration  Func-count       f(x)        Step-size      infinity-norm
     0          24          388.976                            14
     1          72           385.67      0.0292637           16.8  
     2          96           383.54              1           4.15  
     3         120          383.412              1          0.108  
     4         144          383.412              1          0.002  
     5         168          383.412              1        0.00149  
     6         192          383.412              1      6.23e-005  
     7         216          383.412              1      1.01e-005  
   

It is useful when iteration takes long time - I know what's happen
I would appreciate any suggestion

Robert
#
Check out:

https://www.stat.math.ethz.ch/pipermail/r-devel/2006-January/036034.html
On 2/6/06, Robert Mcfadden <robert-mcfadden at o2.pl> wrote:
#
Hi, I never used the function 'optim' so I took a look
at ?optim.
you'll find the following

The 'control' argument is a list that can supply any
of the following components:
     'trace' Non-negative integer. If positive,
tracing information on the progress of the
optimization is produced. Higher values may produce
more tracing information: for method '"L-BFGS-B"'
there are six levels of tracing

so the usage for 'trace' is
where k is >= 0. 
You might want to "play" with different methods and
values for 'trace' and see if you can get what you
want.

hope this helps,

Marco Geraci
--- Robert Mcfadden <robert-mcfadden at o2.pl> wrote:

            
2 days later
#
Hi Bob !

It possible See the first element of controle argument
set it to a positive integer ! 



--- Robert Mcfadden <robert-mcfadden at o2.pl> a ??crit??: