class MagicItemDistribution(object):
# these are the names (and order) of the stats that all magical
# items will have
stats_names = ("dexterity", "constitution", "strength",
"intelligence", "wisdom", "charisma")
def __init__(self, bonus_probs, stats_probs, rso=np.random):
"""Initialize a magic item distribution parameterized by `bonus_probs`
and `stats_probs`.
Parameters
----------
bonus_probs: numpy array of length m
The probabilities of the overall bonuses. Each index in
the array corresponds to the bonus of that amount (e.g.
index 0 is +0, index 1 is +1, etc.)
stats_probs: numpy array of length 6
The probabilities of how the overall bonus is distributed
among the different stats. `stats_probs[i]` corresponds to
the probability of giving a bonus point to the ith stat,
i.e. the value at `MagicItemDistribution.stats_names[i]`.
rso: numpy RandomState object (default: np.random)
The random number generator
"""
# Create the multinomial distributions we'll be using
self.bonus_dist = MultinomialDistribution(bonus_probs, rso=rso)
self.stats_dist = MultinomialDistribution(stats_probs, rso=rso)
def sample(self):
"""Sample a random magical item.
Returns
-------
dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
"""
stats = self._sample_stats()
item_stats = dict(zip(self.stats_names, stats))
return item_stats
def log_pmf(self, item):
"""Compute the log probability the given magical item.
Parameters
----------
item: dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
Returns
-------
float
The value corresponding to log(p(item))
"""
# First pull out the bonus points for each stat, in the
# correct order, then pass that to _stats_log_pmf.
stats = np.array([item[stat] for stat in self.stats_names])
log_pmf = self._stats_log_pmf(stats)
return log_pmf
def pmf(self, item):
"""Compute the probability the given magical item.
Parameters
----------
item: dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
Returns
-------
float
The value corresponding to p(item)
"""
return np.exp(self.log_pmf(item))
def _sample_bonus(self):
"""Sample a value of the overall bonus.
Returns
-------
integer
The overall bonus
"""
# The bonus is essentially just a sample from a multinomial
# distribution with n=1; i.e., only one event occurs.
sample = self.bonus_dist.sample(1)
# `sample` is an array of zeros and a single one at the
# location corresponding to the bonus. We want to convert this
# one into the actual value of the bonus.
bonus = np.argmax(sample)
return bonus
def _sample_stats(self):
"""Sample the overall bonus and how it is distributed across the
different stats.
Returns
-------
numpy array of length 6
The number of bonus points for each stat
"""
# First we need to sample the overall bonus
bonus = self._sample_bonus()
# Then, we use a different multinomial distribution to sample
# how that bonus is distributed. The bonus corresponds to the
# number of events.
stats = self.stats_dist.sample(bonus)
return stats
def _bonus_log_pmf(self, bonus):
"""Evaluate the log-PMF for the given bonus.
Parameters
----------
bonus: integer
The total bonus.
Returns
-------
float
The value corresponding to log(p(bonus))
"""
# Make sure the value that is passed in is within the
# appropriate bounds
if bonus < 0 or bonus >= len(self.bonus_dist.p):
return -np.inf
# Convert the scalar bonus value into a vector of event
# occurrences
x = np.zeros(len(self.bonus_dist.p))
x[bonus] = 1
return self.bonus_dist.log_pmf(x)
def _stats_log_pmf(self, stats):
"""Evaluate the log-PMF for the given distribution of bonus points
across the different stats.
Parameters
----------
stats: numpy array of length 6
The distribution of bonus points across the stats
Returns
-------
float
The value corresponding to log(p(stats))
"""
# There are never any leftover bonus points, so the sum of the
# stats gives us the total bonus.
total_bonus = np.sum(stats)
# First calculate the probability of the total bonus
logp_bonus = self._bonus_log_pmf(total_bonus)
# Then calculate the probability of the stats
logp_stats = self.stats_dist.log_pmf(stats)
# Then multiply them together (using addition, because we are
# working with logs)
log_pmf = logp_bonus + logp_stats
return log_pmf
class MagicItemDistribution(object):
# these are the names (and order) of the stats that all magical
# items will have
stats_names = ("dexterity", "constitution", "strength",
"intelligence", "wisdom", "charisma")
def __init__(self, bonus_probs, stats_probs, rso=np.random):
"""Initialize a magic item distribution parameterized by `bonus_probs`
and `stats_probs`.
Parameters
----------
bonus_probs: numpy array of length m
The probabilities of the overall bonuses. Each index in
the array corresponds to the bonus of that amount (e.g.
index 0 is +0, index 1 is +1, etc.)
stats_probs: numpy array of length 6
The probabilities of how the overall bonus is distributed
among the different stats. `stats_probs[i]` corresponds to
the probability of giving a bonus point to the ith stat,
i.e. the value at `MagicItemDistribution.stats_names[i]`.
rso: numpy RandomState object (default: np.random)
The random number generator
"""
# Create the multinomial distributions we'll be using
self.bonus_dist = MultinomialDistribution(bonus_probs, rso=rso)
self.stats_dist = MultinomialDistribution(stats_probs, rso=rso)
def sample(self):
"""Sample a random magical item.
Returns
-------
dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
"""
stats = self._sample_stats()
item_stats = dict(zip(self.stats_names, stats))
return item_stats
def log_pmf(self, item):
"""Compute the log probability the given magical item.
Parameters
----------
item: dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
Returns
-------
float
The value corresponding to log(p(item))
"""
# First pull out the bonus points for each stat, in the
# correct order, then pass that to _stats_log_pmf.
stats = np.array([item[stat] for stat in self.stats_names])
log_pmf = self._stats_log_pmf(stats)
return log_pmf
def pmf(self, item):
"""Compute the probability the given magical item.
Parameters
----------
item: dictionary
The keys are the names of the stats, and the values are
the bonus conferred to the corresponding stat.
Returns
-------
float
The value corresponding to p(item)
"""
return np.exp(self.log_pmf(item))
def _sample_bonus(self):
"""Sample a value of the overall bonus.
Returns
-------
integer
The overall bonus
"""
# The bonus is essentially just a sample from a multinomial
# distribution with n=1, i.e., only one event occurs.
sample = self.bonus_dist.sample(1)
# `sample` is an array of zeros and a single one at the
# location corresponding to the bonus. We want to convert this
# one into the actual value of the bonus.
bonus = np.argmax(sample)
return bonus
def _sample_stats(self):
"""Sample the overall bonus and how it is distributed across the
different stats.
Returns
-------
numpy array of length 6
The number of bonus points for each stat
"""
# First we need to sample the overall bonus.
bonus = self._sample_bonus()
# Then, we use a different multinomial distribution to sample
# how that bonus is distributed. The bonus corresponds to the
# number of events.
stats = self.stats_dist.sample(bonus)
return stats
def _bonus_log_pmf(self, bonus):
"""Evaluate the log-PMF for the given bonus.
Parameters
----------
bonus: integer
The total bonus.
Returns
-------
float
The value corresponding to log(p(bonus))
"""
# Make sure the value that is passed in is within the
# appropriate bounds.
if bonus < 0 or bonus >= len(self.bonus_dist.p):
return -np.inf
# Convert the scalar bonus value into a vector of event
# occurrences
x = np.zeros(len(self.bonus_dist.p))
x[bonus] = 1
return self.bonus_dist.log_pmf(x)
def _stats_log_pmf(self, stats):
"""Evaluate the log-PMF for the given distribution of bonus points
across the different stats.
Parameters
----------
stats: numpy array of length 6
The distribution of bonus points across the stats
Returns
-------
float
The value corresponding to log(p(stats))
"""
# There are never any leftover bonus points, so the sum of the
# stats gives us the total bonus.
total_bonus = np.sum(stats)
# First calculate the probability of the total bonus
logp_bonus = self._bonus_log_pmf(total_bonus)
# Then calculate the probability of the stats
logp_stats = self.stats_dist.log_pmf(stats)
# Then multiply them together (using addition, because we are
# working in log-space)
log_pmf = logp_bonus + logp_stats
return log_pmf
class MagicItemDistribution(object):
"""rpg游戏中杀死怪物后掉落魔法装备的概率采样"""
stats_names = ('dexterity', 'constitution', 'strength',
'intelligence', 'wisdom', 'charisma')
def __init__(self, bonus_probs, stats_probs, rso=np.random):
"""初始化魔法装备的随机分布
:param bonus_probs: 奖励的概率
:param stats_probs: 奖励在玩家的各个属性中的分布
"""
self.bonus_dist = MultinomialDistribution(bonus_probs, rso=rso)
self.stats_dist = MultinomialDistribution(stats_probs, rso=rso)
def _sample_bonus(self):
"""采样奖励"""
# 只有一个事件发生
sample = self.bonus_dist.sample(1)
bonus = np.argmax(sample)
return bonus
def _sample_stats(self):
"""采样所有的奖励，以及奖励在不同的属性中如何分布"""
bonus = self._sample_bonus()
stats = self.stats_dist.sample(bonus)
return stats
def sample(self):
"""采样获得一个随机的魔法装备"""
stats = self._sample_stats()
item_stats = dict(zip(self.stats_names, stats))
return item_stats
def log_pmf(self, item):
"""计算给定的魔法装备的概率质量函数对数"""
stats = np.array([item[stat] for stat in self.stats_names])
log_pmf = self._stats_log_pmf(stats)
return log_pmf
def pmf(self, item):
"""计算给定装备的概率质量函数"""
return np.exp(self.log_pmf(item))
def _stats_log_pmf(self, stats):
total_bonus = np.sum(stats)
logp_bonus = self._bonus_log_pmf(total_bonus)
logp_stats = self.stats_dist.log_pmf(stats)
log_pmf = logp_bonus + logp_stats
return log_pmf
def _bonus_log_pmf(self, bonus):
if bonus < 0 or bonus >= len(self.bonus_dist.p):
return -np.inf
x = np.zeros(len(self.bonus_dist.p))
x[bonus] = 1
return self.bonus_dist.log_pmf(x)
