API tools faq
paste
LyWang

do_class_AI_1

LyWang
Nov 7th, 2018
153
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Python 7.32 KB | None | 0 0
raw download clone embed print report
  1. import numpy
  2. import random
  3.  
  4.  
  5. class eGreedyAgent:
  6.     value = []
  7.     counter = []
  8.     probability = [0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1]
  9.     score = 0
  10.     preference = 0
  11.     counter_exploration = 0
  12.     counter_exploitation = 0
  13.  
  14.     def __init__(self, e1=0.2, n=10, temp=1):
  15.         self.e = e1
  16.         self.score = 0
  17.         self.value = []
  18.         self.counter = []
  19.         self.temp = temp
  20.         for s in range(n):
  21.             self.value.append([s, 0])
  22.             self.counter.append(0)
  23.  
  24.     def random_choice(self):
  25.         action = random.randint(0, 9)
  26.         self.counter[action] += 1
  27.         return action
  28.  
  29.     def greedy_choice(self):
  30.         r = numpy.random.random()
  31.         max_action = self.max_finder()
  32.  
  33.         if r > self.e:
  34.             self.counter[max_action] += 1
  35.             self.counter_exploitation += 1
  36.             return max_action  # exploit
  37.         else:
  38.             action = random.randint(0, 8)
  39.             if action >= max_action:
  40.                 action += 1
  41.             self.counter[action] += 1
  42.             self.counter_exploration += 1
  43.             return action  # exploration
  44.  
  45.     def persuit_choice(self):
  46.         r = numpy.random.random_sample()
  47.         # print("r is ")
  48.         # print(r)
  49.         max_action = self.max_finder()
  50.         # print('max_action')
  51.         # print(max_action)
  52.         self.probability[max_action] += 0.01 * (1 - self.probability[max_action])
  53.         for u in range(10):
  54.             if u != max_action:
  55.                 self.probability[u] += 0.01 * (0 - self.probability[u])
  56.         possibility_lake = 1
  57.         for i in range(10):
  58.             possibility_lake -= self.probability[i]
  59.             # print('probability is')
  60.             # print(possibility_lake)
  61.             # print(r)
  62.             # print('&&&&')
  63.             # print(possibility_lake)
  64.             # print('')
  65.             if r > possibility_lake:
  66.                 self.counter[i] += 1
  67.                 # print('thus i')
  68.                 # print(i)
  69.                 return i
  70.  
  71.     def exponential_choice(self, temp=1):
  72.         r = numpy.random.random()
  73.         value_exp = []
  74.         maxrange = self.value[0][1]
  75.         for v in range(10):
  76.             if self.value[v][1] > maxrange:
  77.                 maxrange = self.value[v][1]
  78.         for u in range(10):
  79.             if maxrange > 30:
  80.                 value_exp.append(numpy.exp((self.value[u][1] - (maxrange - 30)) / temp))
  81.             elif maxrange < 10:
  82.                 value_exp.append(numpy.exp((self.value[u][1] + (10 - maxrange)) / temp))
  83.             else:
  84.                 value_exp.append(numpy.exp(self.value[u][1] / temp))
  85.  
  86.         # print('*****')
  87.         # print('1')
  88.         # print(self.value)
  89.         # print('2')
  90.         # print(value_exp)
  91.         # print('*****')
  92.         sum_value_exp = sum(value_exp)
  93.         sum_possibility = sum_value_exp
  94.         # print(value_exp)
  95.         for pointer in range(10):
  96.             sum_possibility = sum_possibility - value_exp[pointer]
  97.             # print(sum_possibility / sum_value_exp)
  98.             # print(r)
  99.             # print('-----')
  100.             if sum_possibility / sum_value_exp < r:
  101.                 self.counter[pointer] = self.counter[pointer] + 1
  102.                 # print(pointer)
  103.                 # print('pointer')
  104.                 return pointer
  105.  
  106.     def value_feedback(self, action=0, reward=0):
  107.         self.score += reward
  108.         self.value[action][1] = self.value[action][1] + (reward - self.value[action][1]) / self.counter[action]
  109.  
  110.     def nonstationary_value_feedback(self, action=0, reward=0):
  111.         self.score += reward
  112.         self.value[action][1] = self.value[action][1] + 0.1 * (reward - self.value[action][1])
  113.  
  114.     def reinforcement_comparison_feedback(self, action=0, reward=0):
  115.         # print('reward')
  116.         # print(reward)
  117.         # print('value')
  118.         # print(self.value)
  119.         self.score += reward
  120.         value_exp = []
  121.         maxrange = self.value[0][1]
  122.         for v in range(10):
  123.             if self.value[v][1] > maxrange:
  124.                 maxrange = self.value[v][1]
  125.         # print('maxrange')
  126.         # print(maxrange)
  127.         for u in range(10):
  128.             if maxrange > 30:
  129.                 value_exp.append(numpy.exp((self.value[u][1] - (maxrange - 30)) / self.temp))
  130.             elif maxrange < 10:
  131.                 value_exp.append(numpy.exp((self.value[u][1] + (10 - maxrange)) / self.temp))
  132.             else:
  133.                 value_exp.append(numpy.exp(self.value[u][1] / self.temp))
  134.         self.value[action][1] = self.value[action][1] + 0.1 * (reward - self.preference) * (
  135.                 1 - value_exp[action] / sum(value_exp))
  136.         # print(value_exp)
  137.         # print('aftervalue')
  138.         # print(self.value)
  139.         self.preference = self.preference + 0.1 * (reward - self.preference)
  140.  
  141.     def reinforcement_comparison_feedback_without_factor(self, action=0, reward=0):
  142.         # print('reward')
  143.         # print(reward)
  144.         # print('value')
  145.         # print(self.value)
  146.         self.score += reward
  147.         value_exp = []
  148.         maxrange = self.value[0][1]
  149.         for v in range(10):
  150.             if self.value[v][1] > maxrange:
  151.                 maxrange = self.value[v][1]
  152.         # print('maxrange')
  153.         # print(maxrange)
  154.         for u in range(10):
  155.             if maxrange > 30:
  156.                 value_exp.append(numpy.exp((self.value[u][1] - (maxrange - 30)) / self.temp))
  157.             elif maxrange < 10:
  158.                 value_exp.append(numpy.exp((self.value[u][1] + (10 - maxrange)) / self.temp))
  159.             else:
  160.                 value_exp.append(numpy.exp(self.value[u][1] / self.temp))
  161.         self.value[action][1] = self.value[action][1] + 0.1 * (reward - self.preference)
  162.         # print(value_exp)
  163.         # print('aftervalue')
  164.         # print(self.value)
  165.         self.preference = self.preference + 0.1 * (reward - self.preference)
  166.  
  167.     def max_finder(self):
  168.         max_value = 0
  169.         max_action = 0
  170.         for r1 in range(10):
  171.             if max_value < self.value[r1][1]:
  172.                 max_value = self.value[r1][1]
  173.                 max_action = self.value[r1][0]
  174.         return max_action
  175.  
  176.     def reset(self):
  177.         # print("result")
  178.         # print(self.value)
  179.         # print(self.counter)
  180.         self.score = 0
  181.         self.value = []
  182.         self.counter = []
  183.         self.probability = [0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1]
  184.  
  185.  
  186. class task:
  187.     tasks = []
  188.  
  189.     def __init__(self, num=2000, stationary=True):
  190.         self.tasks = []
  191.         if stationary:
  192.             for n in range(num):
  193.                 subtask = []
  194.                 for j in range(10):
  195.                     subtask.append(numpy.random.rand())
  196.                 self.tasks.append(subtask)
  197.         else:
  198.             for n in range(num):
  199.                 subtask = []
  200.                 for j in range(10):
  201.                     subtask.append(0)
  202.                 self.tasks.append(subtask)
  203.  
  204.     def reward(self, subtask=0, action=0):
  205.         return numpy.random.normal(self.tasks[subtask][action])
  206.  
  207.     def nonstationarymove(self, subtask=0):
  208.         for n in range(10):
  209.             dice = numpy.random.rand()
  210.             if dice > 0.5:
  211.                 self.tasks[subtask][n] += numpy.random.rand()
  212.             else:
  213.                 self.tasks[subtask][n] -= numpy.random.rand()
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!