alperyeg
commented
4 years ago Optimizer Parameters
Crossentropy
- pop_size
- rho: elite fraction
- smoothing: scalar for the distribution
- temp_decay: decay factor
- n_iteration: number of iterations to perform
- distribution: function
- stop_criterion
- seed
Evolution
- seed
- popsize
- CXPB: crossover probability
- MUTPB: mutation probability
- NGEN: number of generations
- indpb: probability of mutation of each element in individual
- tournsize: size of the tournament
- matepar: blending two values during mating
EvolutionStrategies
- learning_rate
- noise_std
- mirrored_sampling_enabled
- fitness_shaping_enabled
- pop_size
- n_iteration
- stop_criterion
- seed
FACE
- min_pop_size
- max_pop_size
- n_elite
- smoothing
- temp_decay
- n_iteration
- distribution
- stop_criterion
- n_expand: amount by which the sample size is increased
Gradient Descent
Classic GD
- learning_rate
- exploration_step_size: standard deviation of random steps used
- n_random_steps: amount of random steps used to estimate gradient
- n_iteration
- stop_criterion
Stochastic GD
- learning_rate
- stochastic_deviation: standard deviation of the random vector used to perturbate the gradient
- stochastic_decay: decay of the influence of the random vector that is added to the gradient
- exploration_step_size
- n_random_steps
- n_iteration
- stop_criterion
Adam
- learning_rate
- exploration_step_size
- n_random_steps
- first_order_decay: amount of decay of the historic first order momentum per gradient descent step
- second_order_decay: amount of decay of the historic second order momentum per gradient descent step
- n_iteration
- stop_criterion
RMSPROP
- learning_rate
- exploration_step_size
- n_random_steps
- momentum_decay
- n_iteration
- stop_criterion
- seed
Gridsearch
- param_grid: (lower_bound, higher_bound, n_steps)
NaturalEvolutionStrategies
- learning_rate_mu
- learning_rate_sigma
- mu
- sigma
- mirrored_sampling_enabled
- fitness_shaping_enabled
- pop_size
- n_iteration
- stop_criterion
- seed
ParallelTempering
- n_parallel_runs: parallel Simulated Annealing runs
- noisy_step: size of the random step
- decay_parameters
- n_iteration
- stop_criterion
- seed
- cooling_schedules
- temperature_bounds
SimulatedAnnealing
- n_parallel_runs
- noisy_step
- temp_decay
- n_iteration
- stop_criterion
- seed
- cooling_schedule
In the optimizer the parameter naming is not standardized due to the namedtuple parameters and the class arguments. There should be an easier way to create parameters.