[SciPy-user] Two questions about maxentropy module
Mon Mar 26 09:04:38 CDT 2007
It's nice to see the scipy.maxentropy module. I have two questions about
Q1) The current model fitting algorithms include CG, BFGS, LBFGSB, Powell,
and Nelder-Mead. How about L-BFGS (unbounded optimization)? Is it
implemented or is there any specific concern of not using it here? Also,
it seems that LBFGSB does not work ...
Q2) bergerexamplesimulated.py example shows how to use bigmodel. However,
it requires the event space to be explicitly enumerated (the enumerated
event space is used when defining a uniform instrumental distribution
for sampling), which is intractable for large event space. E.g., if I
wanna model a maxent distribution over 100 binary variables. What change
should I make to the example code?
Any advice appreciated.
More information about the SciPy-user