We have produced an efficient parallel program running on the DECmpp which 
performs simulated tempering on Ising spin models. However for this program to
run effectively, especially for near zero temperature for optimization problems,
we need to be able to tune the variable parameters in the tempering algorithm.
These parameters are the temperature changes, and the constants which can be 
added to the energy at each temperature. Estimates for these quantities can be 
obtained if the energies are known, however to get reasonable acceptance rates
for the Monte Carlo update on the temperature, these variables will need to be
fine tuned, probably by some dynamical procedure. Good methods of tuning these
variables to obtain good acceptances will need to be found before simulated
tempering can be as effective as simulated annealing for general optimization
problems.