These functions are actually written with C++ and wrapped by pybind11, they are the basic architecture of SA process. User should set the parameter properly, and call the functions if needed.
def setParam(descent_rate,initial_t,final_t,scale,markov_iter,scale_descent_rate)- Set SA parameter.
- descent_rate: descent rate of temperature. After each iteration, the new temperature should be old_temperature*descent_rate
- initial_t: initial temperature
- final_t: final temperature
- scale: scale the range of jumping state
- markov_iter: inner loop of SA, in this loop the temperature won't be change
- scale_descent_rate: descent rate of scale value
def run(show, logger_iter)- Execute the SA flow
- Define whether to display the logger, and the period of iteration to show it.
def showReport()- Show the SA report.
def writeHistory(filename)- Write the data into CSV file.
def plot(filename)- Plot the energy history.
These functions are wrapped into the customized class, and will be called by the C++ SA kernel. In other words, user should define the problem they want to solve in these functions, so that SAAF can run correctly to get the solution.
def jumpState(self,scale,cur_t, iter)- Define how to jump to next state according the current temperature cur_t, scaling ratio scale and iteration iter.
def reverse(self)- Define how to go to the previous state if the current state is rejected.
def storeBest(self)- Store the currently best solution.
def getEnergy(self)->float- Define the energy of the current state. Notice that the state should be store in the attribute of the customized class(by jumpState()), so this function can compute its energy.
def output(self)- This function will be called when the SA process end. User can print some informations or write files to record the result.
def stopCondition(self,final_t,energy,cur_t,iter,ag_r,ab_r,rb_r,best_e)->bool- Define in what condition should the SA process be stopped.
- final_t: final temperature
- energy: current energy
- cur_t: current temperature
- iter: current iteration
- ag_r: rate of accepting good solution(in current temperature)
- ab_r: rate of accepting bad solution
- rb_r: rate of rejecting bad solution
There are 4 demo projects that may help you to understand how this framework works.