Optimizer using Simulated Annealing

SimulatedAnnealingOptimizer

class l2l.optimizers.simulatedannealing.optimizer.SimulatedAnnealingOptimizer(traj, optimizee_create_individual, optimizee_fitness_weights, parameters, optimizee_bounding_func=None)

Bases: l2l.optimizers.optimizer.Optimizer

Class for a generic simulate annealing solver. In the pseudo code the algorithm does:

For n iterations do:

1. Take a step of size noisy step in a random direction
2. If it reduces the cost, keep the solution
3. Otherwise keep with probability exp(- (f_new - f) / T)

NOTE: This expects all parameters of the system to be of floating point

Parameters:

• traj (Trajectory) – Use this trajectory to store the parameters of the specific runs. The parameters should be initialized based on the values in parameters
• optimizee_create_individual – Function that creates a new individual
• optimizee_fitness_weights – Fitness weights. The fitness returned by the Optimizee is multiplied by these values (one for each element of the fitness vector)
• parameters – Instance of namedtuple() SimulatedAnnealingParameters containing the parameters needed by the Optimizer

cooling(temperature, cooling_schedule, temperature_decay, temperature_end, steps_total)

post_process(traj, fitnesses_results)

See post_process()

end(traj)

See end()

SimulatedAnnealingParameters

class l2l.optimizers.simulatedannealing.optimizer.SimulatedAnnealingParameters

Bases: tuple

Parameters:

• n_parallel_runs – Number of individuals per simulation / Number of parallel Simulated Annealing runs
• noisy_step – Size of the random step
• temp_decay – A function of the form f(t) = temperature at time t
• n_iteration – number of iteration to perform
• stop_criterion – Stop if change in fitness is below this value
• seed – Random seed
• cooling_schedule – Which of the available schedules to use

cooling_schedule

n_iteration

n_parallel_runs

noisy_step

seed

stop_criterion

temp_decay