def __init__(
self,
space: Space,
optimizer: Optimizer,
function: Function,
save_agents: Optional[bool] = False,
) -> None:
"""Initialization method.
Args:
space: Space-child instance.
optimizer: Optimizer-child instance.
function: Function or Function-child instance.
save_agents: Saves all agents in the search space.
"""
logger.info("Creating class: Opytimizer.")
# Space
self.space = space
# Optimizer (and its additional variables)
self.optimizer = optimizer
self.optimizer.compile(space)
# Function
self.function = function
# Optimization history
self.history = History(save_agents)
# Current iteration
self.iteration = 0
# Total number of iterations
self.total_iterations = 0
logger.debug(
"Space: %s | Optimizer: %s| Function: %s.",
self.space,
self.optimizer,
self.function,
)
logger.info("Class created.")
def __init__(self, space, optimizer, function, save_agents=False):
"""Initialization method.
Args:
space (Space): Space-child instance.
optimizer (Optimizer): Optimizer-child instance.
function (Function): Function or Function-child instance.
save_agents (bool): Saves all agents in the search space.
"""
logger.info('Creating class: Opytimizer.')
# Space
self.space = space
# Optimizer (and its additional variables)
self.optimizer = optimizer
self.optimizer.compile(space)
# Function
self.function = function
# Optimization history
self.history = History(save_agents)
# Current iteration
self.iteration = 0
# Total number of iterations
self.total_iterations = 0
# Logs the properties
logger.debug('Space: %s | Optimizer: %s| Function: %s.',
self.space, self.optimizer, self.function)
logger.info('Class created.')
class Opytimizer:
"""An Opytimizer class holds all the information needed
in order to perform an optimization task.
"""
def __init__(
self,
space: Space,
optimizer: Optimizer,
function: Function,
save_agents: Optional[bool] = False,
) -> None:
"""Initialization method.
Args:
space: Space-child instance.
optimizer: Optimizer-child instance.
function: Function or Function-child instance.
save_agents: Saves all agents in the search space.
"""
logger.info("Creating class: Opytimizer.")
# Space
self.space = space
# Optimizer (and its additional variables)
self.optimizer = optimizer
self.optimizer.compile(space)
# Function
self.function = function
# Optimization history
self.history = History(save_agents)
# Current iteration
self.iteration = 0
# Total number of iterations
self.total_iterations = 0
logger.debug(
"Space: %s | Optimizer: %s| Function: %s.",
self.space,
self.optimizer,
self.function,
)
logger.info("Class created.")
@property
def space(self) -> Space:
"""Space-child instance (SearchSpace, HyperComplexSpace, etc)."""
return self._space
@space.setter
def space(self, space: Space) -> None:
if not space.built:
raise e.BuildError("`space` should be built before using Opytimizer")
self._space = space
@property
def optimizer(self) -> Optimizer:
"""Optimizer-child instance (PSO, BA, etc)."""
return self._optimizer
@optimizer.setter
def optimizer(self, optimizer: Optimizer) -> None:
if not optimizer.built:
raise e.BuildError("`optimizer` should be built before using Opytimizer")
self._optimizer = optimizer
@property
def function(self) -> Function:
"""Function or Function-child instance (ConstrainedFunction, WeightedFunction, etc)."""
return self._function
@function.setter
def function(self, function: Function) -> None:
if not function.built:
raise e.BuildError("`function` should be built before using Opytimizer")
self._function = function
@property
def history(self) -> History:
"""Optimization history."""
return self._history
@history.setter
def history(self, history: History) -> None:
if not isinstance(history, History):
raise e.TypeError("`history` should be a History")
self._history = history
@property
def iteration(self) -> int:
"""Current iteration."""
return self._iteration
@iteration.setter
def iteration(self, iteration: int) -> None:
if not isinstance(iteration, int):
raise e.TypeError("`iteration` should be an integer")
if iteration < 0:
raise e.ValueError("`iteration` should be >= 0")
self._iteration = iteration
@property
def total_iterations(self) -> int:
"""Total number of iterations."""
return self._total_iterations
@total_iterations.setter
def total_iterations(self, total_iterations: int) -> None:
if not isinstance(total_iterations, int):
raise e.TypeError("`total_iterations` should be an integer")
if total_iterations < 0:
raise e.ValueError("`total_iterations` should be >= 0")
self._total_iterations = total_iterations
@property
def evaluate_args(self) -> List[Any]:
"""Converts the optimizer `evaluate` arguments into real variables.
Returns:
(List[Any]): List of real-attribute variables.
"""
args = signature(self.optimizer.evaluate).parameters
return [getattr(self, v) for v in args]
@property
def update_args(self) -> List[Any]:
"""Converts the optimizer `update` arguments into real variables.
Returns:
(List[Any]): List of real-attribute variables.
"""
args = signature(self.optimizer.update).parameters
return [getattr(self, v) for v in args]
def evaluate(self, callbacks: List[Callback]) -> None:
"""Wraps the `evaluate` pipeline with its corresponding callbacks.
Args:
callbacks: List of callbacks.
"""
# Invokes the `on_evaluate_before` callback
callbacks.on_evaluate_before(*self.evaluate_args)
# Performs an evaluation over the search space
self.optimizer.evaluate(*self.evaluate_args)
# Invokes the `on_evaluate_after` callback
callbacks.on_evaluate_after(*self.evaluate_args)
def update(self, callbacks: List[Callback]) -> None:
"""Wraps the `update` pipeline with its corresponding callbacks.
Args:
callback: List of callbacks.
"""
# Invokes the `on_update_before` callback
callbacks.on_update_before(*self.update_args)
# Performs an update over the search space
self.optimizer.update(*self.update_args)
# Invokes the `on_update_after` callback
callbacks.on_update_after(*self.update_args)
# Regardless of callbacks or not, every update on the search space
# must meet the bounds limits
self.space.clip_by_bound()
def start(
self,
n_iterations: Optional[int] = 1,
callbacks: Optional[List[Callback]] = None,
) -> None:
"""Starts the optimization task.
Args
n_iterations: Maximum number of iterations.
callback: List of callbacks.
"""
logger.info("Starting optimization task.")
# Additional properties
self.n_iterations = n_iterations
callbacks = CallbackVessel(callbacks)
# Triggers starting time
start = time.time()
# Invokes the `on_task_begin` callback
callbacks.on_task_begin(self)
# Evaluates the search space
self.evaluate(callbacks)
# Initializes a progress bar
with tqdm(total=n_iterations, ascii=True) as b:
for t in range(n_iterations):
logger.to_file(f"Iteration {t+1}/{n_iterations}")
# Saves the number of total iterations and current iteration
self.total_iterations += 1
self.iteration = t
# Invokes the `on_iteration_begin` callback
callbacks.on_iteration_begin(self.total_iterations, self)
# Updates the search space
self.update(callbacks)
# Re-evaluates the search space
self.evaluate(callbacks)
# Updates the progress bar status
b.set_postfix(fitness=self.space.best_agent.fit)
b.update()
# Dumps keyword arguments to model's history
self.history.dump(
agents=self.space.agents, best_agent=self.space.best_agent
)
# Invokes the `on_iteration_end` callback
callbacks.on_iteration_end(self.total_iterations, self)
logger.to_file(f"Fitness: {self.space.best_agent.fit}")
logger.to_file(f"Position: {self.space.best_agent.position}")
# Invokes the `on_task_end` callback
callbacks.on_task_end(self)
# Stops the timer and calculates the optimization time
end = time.time()
opt_time = end - start
# Dumps the elapsed time to model's history
self.history.dump(time=opt_time)
logger.info("Optimization task ended.")
logger.info("It took %s seconds.", opt_time)
def save(self, file_path: str) -> None:
"""Saves the optimization model to a dill (pickle) file.
Args:
file_path: Path of file to be saved.
"""
with open(file_path, "wb") as output_file:
dill.dump(self, output_file)
@classmethod
def load(cls, file_path: str) -> None:
"""Loads the optimization model from a dill (pickle) file without needing
to instantiate the class.
Args:
file_path: Path of file to be loaded.
"""
with open(file_path, "rb") as input_file:
opt_model = dill.load(input_file)
return opt_model
if __name__ == '__main__':
# Gathers the input arguments
args = get_arguments()
# Gathering variables from arguments
dataset = args.dataset
descriptor = args.descriptor
fold = args.fold
type = args.type
meta = args.mh
# Defining an input file
input_file = f'output/{meta}_{type}_{dataset}_val_{fold}.pkl'
# Creating a History object
h = History()
# Loading the input file
h.load(input_file)
# Loading the predictions and labels
preds, y = l.load_candidates(dataset, 'test', fold)
# If descriptor is global-based
if descriptor == 'global':
# Gets the global predictors
preds = preds[:, :35]
# If descriptor is cnn-based
elif descriptor == 'cnn':
# Gets the CNN predictors
from opytimizer.utils.history import History # File name to be loaded file_name = '' # Creating an empty History object h = History() # Loading history from pickle file h.load(file_name) # Displaying content print(h)
import numpy as np
import opytimizer.visualization.convergence as c
from opytimizer.utils.history import History
# Creating the history object
history = History()
# Loading saved optimization task
history.load('')
# Gathering desired keys from the object
# In this case, we will the first agent's position and fitness
agent_pos = history.get(key='agents', index=(0, 0))
agent_fit = history.get(key='agents', index=(0, 1))
# We will also gather the best agent's position and fitness
best_agent_pos = history.get(key='best_agent', index=(0,))
best_agent_fit = history.get(key='best_agent', index=(1,))
# Plotting convergence graphs
# Plotting the convergence of agent's positions
c.plot(agent_pos[0], agent_pos[1], labels=['$x_0$', '$x_1$'],
title='Sphere Function: $x^2 \mid x \in [-10, 10]$', subtitle='Agent: 0 | Algorithm: Particle Swarm Optimization')
# Plotting the convergence of best agent's positions
c.plot(best_agent_pos[0], best_agent_pos[1], labels=['$x^*_0$', '$x^*_1$'],
title='Sphere Function: $x^2 \mid x \in [-10, 10]$', subtitle="Agent: Best | Algorithm: Particle Swarm Optimization")
# Plotting the convergence of agent's and best agent's fitness
c.plot(agent_fit, best_agent_fit, labels=[
from opytimizer.utils.history import History # Instantiates the History h = History() # Dumps a variable (it will be converted into a list) h.dump(x=1) h.dump(x=2) h.dump(x=3) # Any variable will be converted into a list # Even lists, dictionaries, etc h.dump(y=[1]) # Access the variables print(h.x) print(h.y)
class Opytimizer:
"""An Opytimizer class holds all the information needed
in order to perform an optimization task.
"""
def __init__(self, space, optimizer, function, save_agents=False):
"""Initialization method.
Args:
space (Space): Space-child instance.
optimizer (Optimizer): Optimizer-child instance.
function (Function): Function or Function-child instance.
save_agents (bool): Saves all agents in the search space.
"""
logger.info('Creating class: Opytimizer.')
# Space
self.space = space
# Optimizer (and its additional variables)
self.optimizer = optimizer
self.optimizer.compile(space)
# Function
self.function = function
# Optimization history
self.history = History(save_agents)
# Current iteration
self.iteration = 0
# Total number of iterations
self.total_iterations = 0
# Logs the properties
logger.debug('Space: %s | Optimizer: %s| Function: %s.',
self.space, self.optimizer, self.function)
logger.info('Class created.')
@property
def space(self):
"""Space: Space-child instance (SearchSpace, HyperComplexSpace, etc).
"""
return self._space
@space.setter
def space(self, space):
if not space.built:
raise e.BuildError('`space` should be built before using Opytimizer')
self._space = space
@property
def optimizer(self):
"""Optimizer: Optimizer-child instance (PSO, BA, etc).
"""
return self._optimizer
@optimizer.setter
def optimizer(self, optimizer):
if not optimizer.built:
raise e.BuildError('`optimizer` should be built before using Opytimizer')
self._optimizer = optimizer
@property
def function(self):
"""Function: Function or Function-child instance (ConstrainedFunction, WeightedFunction, etc).
"""
return self._function
@function.setter
def function(self, function):
if not function.built:
raise e.BuildError('`function` should be built before using Opytimizer')
self._function = function
@property
def history(self):
"""History: Optimization history.
"""
return self._history
@history.setter
def history(self, history):
if not isinstance(history, History):
raise e.TypeError('`history` should be a History')
self._history = history
@property
def iteration(self):
"""int: Current iteration.
"""
return self._iteration
@iteration.setter
def iteration(self, iteration):
if not isinstance(iteration, int):
raise e.TypeError('`iteration` should be an integer')
if iteration < 0:
raise e.ValueError('`iteration` should be >= 0')
self._iteration = iteration
@property
def total_iterations(self):
"""int: Total number of iterations.
"""
return self._total_iterations
@total_iterations.setter
def total_iterations(self, total_iterations):
if not isinstance(total_iterations, int):
raise e.TypeError('`total_iterations` should be an integer')
if total_iterations < 0:
raise e.ValueError('`total_iterations` should be >= 0')
self._total_iterations = total_iterations
@property
def evaluate_args(self):
"""Converts the optimizer `evaluate` arguments into real variables.
"""
# Inspects the `evaluate` and retrieves its parameters
args = signature(self.optimizer.evaluate).parameters
return [getattr(self, v) for v in args]
@property
def update_args(self):
"""Converts the optimizer `update` arguments into real variables.
"""
# Inspects the `update` and retrieves its parameters
args = signature(self.optimizer.update).parameters
return [getattr(self, v) for v in args]
def evaluate(self, callbacks):
"""Wraps the `evaluate` pipeline with its corresponding callbacks.
Args:
callback (list): List of callbacks.
"""
# Invokes the `on_evaluate_before` callback
callbacks.on_evaluate_before(*self.evaluate_args)
# Performs an evaluation over the search space
self.optimizer.evaluate(*self.evaluate_args)
# Invokes the `on_evaluate_after` callback
callbacks.on_evaluate_after(*self.evaluate_args)
def update(self, callbacks):
"""Wraps the `update` pipeline with its corresponding callbacks.
Args:
callback (list): List of callbacks.
"""
# Invokes the `on_update_before` callback
callbacks.on_update_before(*self.update_args)
# Performs an update over the search space
self.optimizer.update(*self.update_args)
# Invokes the `on_update_after` callback
callbacks.on_update_after(*self.update_args)
# Regardless of callbacks or not, every update on the search space
# must meet the bounds limits
self.space.clip_by_bound()
def start(self, n_iterations=1, callbacks=None):
"""Starts the optimization task.
Args
n_iterations (int): Maximum number of iterations.
callback (list): List of callbacks.
"""
logger.info('Starting optimization task.')
# Additional properties
self.n_iterations = n_iterations
callbacks = CallbackVessel(callbacks)
# Triggers starting time
start = time.time()
# Evaluates the search space
self.evaluate(callbacks)
# Initializes a progress bar
with tqdm(total=n_iterations) as b:
# Loops through all iterations
for t in range(n_iterations):
logger.to_file(f'Iteration {t+1}/{n_iterations}')
# Saves the number of total iterations and current iteration
self.total_iterations += 1
self.iteration = t
# Invokes the `on_iteration_begin` callback
callbacks.on_iteration_begin(self.total_iterations, self)
# Updates the search space
self.update(callbacks)
# Re-evaluates the search space
self.evaluate(callbacks)
# Updates the progress bar status
b.set_postfix(fitness=self.space.best_agent.fit)
b.update()
# Dumps keyword arguments to model's history
self.history.dump(agents=self.space.agents,
best_agent=self.space.best_agent)
# Invokes the `on_iteration_end` callback
callbacks.on_iteration_end(self.total_iterations, self)
logger.to_file(f'Fitness: {self.space.best_agent.fit}')
logger.to_file(f'Position: {self.space.best_agent.position}')
# Stops the timer and calculates the optimization time
end = time.time()
opt_time = end - start
# Dumps the elapsed time to model's history
self.history.dump(time=opt_time)
logger.info('Optimization task ended.')
logger.info('It took %s seconds.', opt_time)
def save(self, file_path):
"""Saves the optimization model to a pickle file.
Args:
file_path (str): Path of file to be saved.
"""
# Opens an output file
with open(file_path, 'wb') as output_file:
# Dumps object to file
pickle.dump(self, output_file)
@classmethod
def load(cls, file_path):
"""Loads the optimization model from a pickle file without needing
to instantiate the class.
Args:
file_path (str): Path of file to be loaded.
"""
# Opens an input file
with open(file_path, "rb") as input_file:
# Loads object from file
opt_model = pickle.load(input_file)
return opt_model