# -*- coding: utf-8 -*-
"""The Simulator takes in a :obj:`seagull.Board`, and runs a simulation given a
set number of iterations and a rule. For each iteration, the rule is applied to
the Board in order to evolve the lifeforms. After the simulation, run
statistics are returned.
.. code-block:: python
import seagull as sg
board = sg.Board()
board.add(Blinker(), loc=(0,0))
# Initialize a simulator
sim = sg.Simulator(board)
stats = sim.run(sg.rules.conway_classic, iters=1000)
You can always get the history of the whole simulation by calling the
`get_history()` method. The length of the history will always be equal to
:code:`iters + 1` since we include the initial state
.. note::
Running a simulation does not change the :code:`state` attribute of the
board. Internally, the simulator makes a copy of that layout and updates
that instead. This is to avoid unintended behaviour when running
simulations again and again.
Various statistics such as entropy, peak cell coverage, and the like are
returned as a dictionary. This gives us an idea on the characteristics of the
simulation experiment.
.. note::
Some statistics are highly-dependent on the size of the board and the
number of iterations. For example, peak cell coverage (pertaining to the
max. amount of active cells during the whole run) depends on board size. If
you have better ideas for computing these statistics, please open-up an
Issue!
The :code:`run()` method only computes the progress of the board for the whole
simulation, but it does not animate it yet. To create an animation, call the
:code:`animate()` method:
.. code-block:: python
sim.animate()
This returns a :obj:`matplotlib.animation.FuncAnimation` that you can turn into
an interactive animation in your notebook or exported as a GIF.
.. note::
When exporting to GIF, it is required to have the ffmpeg backend installed.
"""
# Import standard library
from typing import Callable, Union
# Import modules
import matplotlib.pyplot as plt
import numpy as np
from loguru import logger
from matplotlib import animation
from .board import Board
from .utils import statistics as stats
[docs]class Simulator:
[docs] def __init__(self, board: Board):
"""Initialize the class
Parameters
----------
board : seagull.Board
The board to run the simulation on
"""
self.board = board
self.history = [] # type: list
self.stats = {} # type: dict
[docs] def run(self, rule: Callable, iters: int, **kwargs) -> dict:
"""Run the simulation for a given number of iterations
Parameters
----------
rule : callable
Callable that takes in an array and returns an array of the same
shape.
iters : int
Number of iterations to run the simulation.
Returns
-------
dict
Computed statistics for the simulation run
"""
layout = self.board.state.copy()
# Append the initial state
self.history.append(layout)
# Run simulation
for i in range(iters):
layout = rule(layout, **kwargs)
self.history.append(layout)
self.stats = self.compute_statistics(self.get_history())
return self.stats
[docs] def compute_statistics(self, history: Union[list, np.ndarray]) -> dict:
"""Compute various statistics for the board
Parameters
----------
history : list or numpy.ndarray
The simulation history
Returns
-------
dict
Compute statistics
"""
logger.info("Computing simulation statistics...")
sim_stats = {
"peak_cell_coverage": np.max(
[stats.cell_coverage(h) for h in history]
),
"avg_cell_coverage": np.mean(
[stats.cell_coverage(h) for h in history]
),
"avg_shannon_entropy": np.mean(
[stats.shannon_entropy(h) for h in history]
),
"peak_shannon_entropy": np.max(
[stats.shannon_entropy(h) for h in history]
),
}
return sim_stats
[docs] def get_history(self, exclude_init=False) -> np.ndarray:
"""Get the simulation history
Parameters
----------
exclude_init: bool
If True, then excludes the initial state in the history
Returns
-------
numpy.ndarray
Simulation history of shape :code:`(iters+1, board.size[0],
board.size[1])`
"""
history = self.history[1:] if exclude_init else self.history
return np.asarray(history)
[docs] def animate(self, figsize=(5, 5), interval=100) -> animation.FuncAnimation:
"""Animate the resulting simulation
Parameters
----------
figsize : tuple
Size of the output figure
interval : int
Interval for transitioning between frames
Returns
-------
matplotlib.animation.FuncAnimation
Animation generated from the run
"""
if not self.history:
msg = "The run() argument must be executed first"
logger.error(msg)
raise ValueError(msg)
logger.info("Rendering animation...")
fig = plt.figure(figsize=figsize)
ax = fig.add_axes([0, 0, 1, 1], xticks=[], yticks=[], frameon=False)
X_blank = np.zeros(self.board.size, dtype=bool)
im = ax.imshow(X_blank, cmap=plt.cm.binary, interpolation="nearest")
im.set_clim(-0.05, 1)
def _animate(i, history):
current_pos = history[i]
im.set_data(current_pos)
return (im,)
def _init():
im.set_data(X_blank)
return (im,)
history = self.get_history()
anim = animation.FuncAnimation(
fig,
func=_animate,
frames=range(history.shape[0]),
init_func=_init,
interval=interval,
fargs=(history,),
blit=True,
)
return anim