API tools faq
paste
Advertisement
Guest User

normalizacija

a guest
May 21st, 2018
82
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 11.46 KB | None | 0 0
raw download clone embed print report
  1.  
  2. from math import exp
  3. import random
  4.  
  5.  
  6. def initialize_network(n_inputs, n_hidden, n_outputs):
  7. network = list()
  8. hidden_layer = [{'weights': [random.random() for i in range(n_inputs + 1)]} for i in range(n_hidden)]
  9. network.append(hidden_layer)
  10. output_layer = [{'weights': [random.random() for i in range(n_hidden + 1)]} for i in range(n_outputs)]
  11. network.append(output_layer)
  12. return network
  13.  
  14.  
  15. def activate(weights, inputs):
  16. activation = weights[-1]
  17. for i in range(len(weights) - 1):
  18. activation += weights[i] * inputs[i]
  19. return activation
  20.  
  21.  
  22. def transfer(activation):
  23. return 1.0 / (1.0 + exp(-activation))
  24.  
  25.  
  26. def forward_propagate(network, row):
  27. inputs = row
  28. for layer in network:
  29. new_inputs = []
  30. for neuron in layer:
  31. activation = activate(neuron['weights'], inputs)
  32. neuron['output'] = transfer(activation)
  33. new_inputs.append(neuron['output'])
  34. inputs = new_inputs
  35. return inputs
  36.  
  37.  
  38. def transfer_derivative(output):
  39. return output * (1.0 - output)
  40.  
  41.  
  42. def backward_propagate_error(network, expected):
  43. for i in reversed(range(len(network))):
  44. layer = network[i]
  45. errors = list()
  46. if i != len(network) - 1:
  47. for j in range(len(layer)):
  48. error = 0.0
  49. for neuron in network[i + 1]:
  50. error += (neuron['weights'][j] * neuron['delta'])
  51. errors.append(error)
  52. else:
  53. for j in range(len(layer)):
  54. neuron = layer[j]
  55. errors.append(expected[j] - neuron['output'])
  56. for j in range(len(layer)):
  57. neuron = layer[j]
  58. neuron['delta'] = errors[j] * transfer_derivative(neuron['output'])
  59.  
  60.  
  61. def update_weights(network, row, l_rate):
  62. for i in range(len(network)):
  63. inputs = row[:-1]
  64. if i != 0:
  65. inputs = [neuron['output'] for neuron in network[i - 1]]
  66. for neuron in network[i]:
  67. for j in range(len(inputs)):
  68. neuron['weights'][j] += l_rate * neuron['delta'] * inputs[j]
  69. neuron['weights'][-1] += l_rate * neuron['delta']
  70.  
  71.  
  72. def train_network(network, train, l_rate, n_epoch, n_outputs):
  73. for epoch in range(n_epoch):
  74. sum_error = 0
  75. for row in train:
  76. outputs = forward_propagate(network, row)
  77. expected = [0 for i in range(n_outputs)]
  78. expected[row[-1]] = 1
  79. sum_error += sum([(expected[i] - outputs[i]) ** 2 for i in range(len(expected))])
  80. backward_propagate_error(network, expected)
  81. update_weights(network, row, l_rate)
  82. # print('>epoch=%d, lrate=%.3f, error=%.3f' % (epoch, l_rate, sum_error))
  83.  
  84.  
  85. def predict(network, row):
  86. outputs = forward_propagate(network, row)
  87. return outputs.index(max(outputs))
  88.  
  89.  
  90. random.seed(1)
  91. training_data = [
  92. [3.6216, 8.6661, -2.8073, -0.44699, 0],
  93. [4.5459, 8.1674, -2.4586, -1.4621, 0],
  94. [3.866, -2.6383, 1.9242, 0.10645, 0],
  95. [3.4566, 9.5228, -4.0112, -3.5944, 0],
  96. [0.32924, -4.4552, 4.5718, -0.9888, 0],
  97. [4.3684, 9.6718, -3.9606, -3.1625, 0],
  98. [3.5912, 3.0129, 0.72888, 0.56421, 0],
  99. [2.0922, -6.81, 8.4636, -0.60216, 0],
  100. [3.2032, 5.7588, -0.75345, -0.61251, 0],
  101. [1.5356, 9.1772, -2.2718, -0.73535, 0],
  102. [1.2247, 8.7779, -2.2135, -0.80647, 0],
  103. [3.9899, -2.7066, 2.3946, 0.86291, 0],
  104. [1.8993, 7.6625, 0.15394, -3.1108, 0],
  105. [-1.5768, 10.843, 2.5462, -2.9362, 0],
  106. [3.404, 8.7261, -2.9915, -0.57242, 0],
  107. [4.6765, -3.3895, 3.4896, 1.4771, 0],
  108. [2.6719, 3.0646, 0.37158, 0.58619, 0],
  109. [0.80355, 2.8473, 4.3439, 0.6017, 0],
  110. [1.4479, -4.8794, 8.3428, -2.1086, 0],
  111. [5.2423, 11.0272, -4.353, -4.1013, 0],
  112. [5.7867, 7.8902, -2.6196, -0.48708, 0],
  113. [0.3292, -4.4552, 4.5718, -0.9888, 0],
  114. [3.9362, 10.1622, -3.8235, -4.0172, 0],
  115. [0.93584, 8.8855, -1.6831, -1.6599, 0],
  116. [4.4338, 9.887, -4.6795, -3.7483, 0],
  117. [0.7057, -5.4981, 8.3368, -2.8715, 0],
  118. [1.1432, -3.7413, 5.5777, -0.63578, 0],
  119. [-0.38214, 8.3909, 2.1624, -3.7405, 0],
  120. [6.5633, 9.8187, -4.4113, -3.2258, 0],
  121. [4.8906, -3.3584, 3.4202, 1.0905, 0],
  122. [-0.24811, -0.17797, 4.9068, 0.15429, 0],
  123. [1.4884, 3.6274, 3.308, 0.48921, 0],
  124. [4.2969, 7.617, -2.3874, -0.96164, 0],
  125. [-0.96511, 9.4111, 1.7305, -4.8629, 0],
  126. [-1.6162, 0.80908, 8.1628, 0.60817, 0],
  127. [2.4391, 6.4417, -0.80743, -0.69139, 0],
  128. [2.6881, 6.0195, -0.46641, -0.69268, 0],
  129. [3.6289, 0.81322, 1.6277, 0.77627, 0],
  130. [4.5679, 3.1929, -2.1055, 0.29653, 0],
  131. [3.4805, 9.7008, -3.7541, -3.4379, 0],
  132. [4.1711, 8.722, -3.0224, -0.59699, 0],
  133. [-0.2062, 9.2207, -3.7044, -6.8103, 0],
  134. [-0.0068919, 9.2931, -0.41243, -1.9638, 0],
  135. [0.96441, 5.8395, 2.3235, 0.066365, 0],
  136. [2.8561, 6.9176, -0.79372, 0.48403, 0],
  137. [-0.7869, 9.5663, -3.7867, -7.5034, 0],
  138. [2.0843, 6.6258, 0.48382, -2.2134, 0],
  139. [-0.7869, 9.5663, -3.7867, -7.5034, 0],
  140. [3.9102, 6.065, -2.4534, -0.68234, 0],
  141. [1.6349, 3.286, 2.8753, 0.087054, 0],
  142. [4.3239, -4.8835, 3.4356, -0.5776, 0],
  143. [5.262, 3.9834, -1.5572, 1.0103, 0],
  144. [3.1452, 5.825, -0.51439, -1.4944, 0],
  145. [2.549, 6.1499, -1.1605, -1.2371, 0],
  146. [4.9264, 5.496, -2.4774, -0.50648, 0],
  147. [4.8265, 0.80287, 1.6371, 1.1875, 0],
  148. [2.5635, 6.7769, -0.61979, 0.38576, 0],
  149. [5.807, 5.0097, -2.2384, 0.43878, 0],
  150. [3.1377, -4.1096, 4.5701, 0.98963, 0],
  151. [-0.78289, 11.3603, -0.37644, -7.0495, 0],
  152. [-1.3971, 3.3191, -1.3927, -1.9948, 1],
  153. [0.39012, -0.14279, -0.031994, 0.35084, 1],
  154. [-1.6677, -7.1535, 7.8929, 0.96765, 1],
  155. [-3.8483, -12.8047, 15.6824, -1.281, 1],
  156. [-3.5681, -8.213, 10.083, 0.96765, 1],
  157. [-2.2804, -0.30626, 1.3347, 1.3763, 1],
  158. [-1.7582, 2.7397, -2.5323, -2.234, 1],
  159. [-0.89409, 3.1991, -1.8219, -2.9452, 1],
  160. [0.3434, 0.12415, -0.28733, 0.14654, 1],
  161. [-0.9854, -6.661, 5.8245, 0.5461, 1],
  162. [-2.4115, -9.1359, 9.3444, -0.65259, 1],
  163. [-1.5252, -6.2534, 5.3524, 0.59912, 1],
  164. [-0.61442, -0.091058, -0.31818, 0.50214, 1],
  165. [-0.36506, 2.8928, -3.6461, -3.0603, 1],
  166. [-5.9034, 6.5679, 0.67661, -6.6797, 1],
  167. [-1.8215, 2.7521, -0.72261, -2.353, 1],
  168. [-0.77461, -1.8768, 2.4023, 1.1319, 1],
  169. [-1.8187, -9.0366, 9.0162, -0.12243, 1],
  170. [-3.5801, -12.9309, 13.1779, -2.5677, 1],
  171. [-1.8219, -6.8824, 5.4681, 0.057313, 1],
  172. [-0.3481, -0.38696, -0.47841, 0.62627, 1],
  173. [0.47368, 3.3605, -4.5064, -4.0431, 1],
  174. [-3.4083, 4.8587, -0.76888, -4.8668, 1],
  175. [-1.6662, -0.30005, 1.4238, 0.024986, 1],
  176. [-2.0962, -7.1059, 6.6188, -0.33708, 1],
  177. [-2.6685, -10.4519, 9.1139, -1.7323, 1],
  178. [-0.47465, -4.3496, 1.9901, 0.7517, 1],
  179. [1.0552, 1.1857, -2.6411, 0.11033, 1],
  180. [1.1644, 3.8095, -4.9408, -4.0909, 1],
  181. [-4.4779, 7.3708, -0.31218, -6.7754, 1],
  182. [-2.7338, 0.45523, 2.4391, 0.21766, 1],
  183. [-2.286, -5.4484, 5.8039, 0.88231, 1],
  184. [-1.6244, -6.3444, 4.6575, 0.16981, 1],
  185. [0.50813, 0.47799, -1.9804, 0.57714, 1],
  186. [1.6408, 4.2503, -4.9023, -2.6621, 1],
  187. [0.81583, 4.84, -5.2613, -6.0823, 1],
  188. [-5.4901, 9.1048, -0.38758, -5.9763, 1],
  189. [-3.2238, 2.7935, 0.32274, -0.86078, 1],
  190. [-2.0631, -1.5147, 1.219, 0.44524, 1],
  191. [-0.91318, -2.0113, -0.19565, 0.066365, 1],
  192. [0.6005, 1.9327, -3.2888, -0.32415, 1],
  193. [0.91315, 3.3377, -4.0557, -1.6741, 1],
  194. [-0.28015, 3.0729, -3.3857, -2.9155, 1],
  195. [-3.6085, 3.3253, -0.51954, -3.5737, 1],
  196. [-6.2003, 8.6806, 0.0091344, -3.703, 1],
  197. [-4.2932, 3.3419, 0.77258, -0.99785, 1],
  198. [-3.0265, -0.062088, 0.68604, -0.055186, 1],
  199. [-1.7015, -0.010356, -0.99337, -0.53104, 1],
  200. [-0.64326, 2.4748, -2.9452, -1.0276, 1],
  201. [-0.86339, 1.9348, -2.3729, -1.0897, 1],
  202. [-2.0659, 1.0512, -0.46298, -1.0974, 1],
  203. [-2.1333, 1.5685, -0.084261, -1.7453, 1],
  204. [-1.2568, -1.4733, 2.8718, 0.44653, 1],
  205. [-3.1128, -6.841, 10.7402, -1.0172, 1],
  206. [-4.8554, -5.9037, 10.9818, -0.82199, 1],
  207. [-2.588, 3.8654, -0.3336, -1.2797, 1],
  208. [0.24394, 1.4733, -1.4192, -0.58535, 1],
  209. [-1.5322, -5.0966, 6.6779, 0.17498, 1],
  210. [-4.0025, -13.4979, 17.6772, -3.3202, 1],
  211. [-4.0173, -8.3123, 12.4547, -1.4375, 1]
  212. ]
  213.  
  214. testing_data = [
  215. [2.888, 0.44696, 4.5907, -0.24398, 0],
  216. [0.49665, 5.527, 1.7785, -0.47156, 0],
  217. [4.2586, 11.2962, -4.0943, -4.3457, 0],
  218. [1.7939, -1.1174, 1.5454, -0.26079, 0],
  219. [5.4021, 3.1039, -1.1536, 1.5651, 0],
  220. [2.5367, 2.599, 2.0938, 0.20085, 0],
  221. [4.6054, -4.0765, 2.7587, 0.31981, 0],
  222. [-1.979, 3.2301, -1.3575, -2.5819, 1],
  223. [-0.4294, -0.14693, 0.044265, -0.15605, 1],
  224. [-2.234, -7.0314, 7.4936, 0.61334, 1],
  225. [-4.211, -12.4736, 14.9704, -1.3884, 1],
  226. [-3.8073, -8.0971, 10.1772, 0.65084, 1],
  227. [-2.5912, -0.10554, 1.2798, 1.0414, 1],
  228. [-2.2482, 3.0915, -2.3969, -2.6711, 1]
  229. ]
  230.  
  231.  
  232. # 0 vistinska, 1 falsifikat
  233.  
  234. # f-ja za normalizacija
  235. def minMaxNrom(training_data, testing_data):
  236. # vashiot kod tuka
  237. new_data = [[] for i in range(len(training_data))] # lista od listi, za da mozheme da gi dodavame novite elem. kolona po kolona
  238. new_testing_data = [[] for i in range(len(testing_data))] # isto |
  239.  
  240. for kolona in range(0, len(training_data[0]) - 1):
  241. min = 1000
  242. max = -1000
  243. for row in training_data:
  244. if row[kolona] > max:
  245. max = row[kolona]
  246. if row[kolona] < min:
  247. min = row[kolona]
  248. # print(min)
  249. # print(max)
  250. # print
  251. index = 0
  252. for row in training_data:
  253. nov = (row[kolona] - min) / (max - min)
  254. new_data[index].append(nov) # se dodavaat 1 po drug na ista adresa
  255. index += 1
  256. # dodavanje na odgovorite, t.e. tochnite klasifikacii na kraj
  257. index = -1
  258. for row in new_data:
  259. index += 1
  260. row.append(training_data[index][len(training_data[index]) - 1])
  261.  
  262. for kolona in range(0, len(testing_data[0]) - 1):
  263. min = 1000
  264. max = -1000
  265. for row in testing_data:
  266. if row[kolona] > max:
  267. max = row[kolona]
  268. if row[kolona] < min:
  269. min = row[kolona]
  270. # print(min)
  271. # print(max)
  272. # print
  273. index = 0
  274. for row in testing_data:
  275. nov = (row[kolona] - min) / (max - min)
  276. new_testing_data[index].append(nov) # se dodavaat 1 po drug na ista adresa
  277. index += 1
  278. index = -1
  279. for row in new_testing_data:
  280. index += 1
  281. row.append(testing_data[index][len(testing_data[index]) - 1])
  282.  
  283. return new_data, new_testing_data
  284.  
  285.  
  286. if __name__ == "__main__":
  287. # vashiot kod tuka
  288. neural_network1 = initialize_network(4, 3, 2)
  289. neural_network2 = initialize_network(4, 3, 2)
  290. normalized_training_data, normalized_testing_data = minMaxNrom(training_data, testing_data)
  291. for row in normalized_testing_data:
  292. print(row)
  293. train_network(neural_network1, training_data, 0.3, 50, 2)
  294. train_network(neural_network2, normalized_training_data, 0.3, 50, 2)
  295.  
  296. normal = 0
  297. for row in testing_data:
  298. predvideno = predict(neural_network1, row)
  299. if predvideno == row[-1]:
  300. normal += 1
  301. print("Prva mrezha:")
  302. print(normal)
  303.  
  304. normalized = 0
  305. for row in normalized_testing_data:
  306. predvideno = predict(neural_network2, row)
  307. if predvideno == row[-1]:
  308. normalized += 1
  309. print("Vtora mrezha:")
  310. print(normalized)
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
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!