kerasで九九

概要

kerasで九九、やってみた。

結果

81/81 [==============================] - 0s - loss: 0.0277 - val_loss: 0.0276
Epoch 858/1000
81/81 [==============================] - 0s - loss: 0.0276 - val_loss: 0.0275
Epoch 859/1000
81/81 [==============================] - 0s - loss: 0.0276 - val_loss: 0.0275
Epoch 860/1000
81/81 [==============================] - 0s - loss: 0.0276 - val_loss: 0.0275
Epoch 861/1000
81/81 [==============================] - 0s - loss: 0.0276 - val_loss: 0.0274
Epoch 862/1000
81/81 [==============================] - 0s - loss: 0.0273 - val_loss: 0.0273
Epoch 863/1000
81/81 [==============================] - 0s - loss: 0.0273 - val_loss: 0.0273
Epoch 864/1000
81/81 [==============================] - 0s - loss: 0.0274 - val_loss: 0.0273
Epoch 865/1000
81/81 [==============================] - 0s - loss: 0.0273 - val_loss: 0.0272
Epoch 866/1000
81/81 [==============================] - 0s - loss: 0.0272 - val_loss: 0.0272
Epoch 867/1000
81/81 [==============================] - 0s - loss: 0.0272 - val_loss: 0.0273
Epoch 868/1000
81/81 [==============================] - 0s - loss: 0.0273 - val_loss: 0.0272
Epoch 869/1000
81/81 [==============================] - 0s - loss: 0.0272 - val_loss: 0.0270
Epoch 870/1000
81/81 [==============================] - 0s - loss: 0.0271 - val_loss: 0.0269
Epoch 871/1000
81/81 [==============================] - 0s - loss: 0.0269 - val_loss: 0.0268
Epoch 872/1000
81/81 [==============================] - 0s - loss: 0.0269 - val_loss: 0.0268
Epoch 873/1000
81/81 [==============================] - 0s - loss: 0.0269 - val_loss: 0.0267
Epoch 874/1000
81/81 [==============================] - 0s - loss: 0.0268 - val_loss: 0.0266
Epoch 875/1000
81/81 [==============================] - 0s - loss: 0.0267 - val_loss: 0.0265
Epoch 876/1000
81/81 [==============================] - 0s - loss: 0.0266 - val_loss: 0.0264
Epoch 877/1000
81/81 [==============================] - 0s - loss: 0.0265 - val_loss: 0.0264
Epoch 878/1000
81/81 [==============================] - 0s - loss: 0.0264 - val_loss: 0.0263
Epoch 879/1000
81/81 [==============================] - 0s - loss: 0.0264 - val_loss: 0.0263
Epoch 880/1000
81/81 [==============================] - 0s - loss: 0.0263 - val_loss: 0.0262
Epoch 881/1000
81/81 [==============================] - 0s - loss: 0.0262 - val_loss: 0.0261
Epoch 882/1000
81/81 [==============================] - 0s - loss: 0.0262 - val_loss: 0.0260
Epoch 883/1000
81/81 [==============================] - 0s - loss: 0.0261 - val_loss: 0.0260
Epoch 884/1000
81/81 [==============================] - 0s - loss: 0.0260 - val_loss: 0.0260
Epoch 885/1000
81/81 [==============================] - 0s - loss: 0.0260 - val_loss: 0.0259
Epoch 886/1000
81/81 [==============================] - 0s - loss: 0.0260 - val_loss: 0.0259
Epoch 887/1000
81/81 [==============================] - 0s - loss: 0.0260 - val_loss: 0.0258
Epoch 888/1000
81/81 [==============================] - 0s - loss: 0.0258 - val_loss: 0.0257
Epoch 889/1000
81/81 [==============================] - 0s - loss: 0.0257 - val_loss: 0.0257
Epoch 890/1000
81/81 [==============================] - 0s - loss: 0.0258 - val_loss: 0.0257
Epoch 891/1000
81/81 [==============================] - 0s - loss: 0.0258 - val_loss: 0.0256
Epoch 892/1000
81/81 [==============================] - 0s - loss: 0.0258 - val_loss: 0.0255
Epoch 893/1000
81/81 [==============================] - 0s - loss: 0.0256 - val_loss: 0.0255
Epoch 894/1000
81/81 [==============================] - 0s - loss: 0.0255 - val_loss: 0.0255
Epoch 895/1000
81/81 [==============================] - 0s - loss: 0.0255 - val_loss: 0.0256
Epoch 896/1000
81/81 [==============================] - 0s - loss: 0.0257 - val_loss: 0.0256
Epoch 897/1000
81/81 [==============================] - 0s - loss: 0.0256 - val_loss: 0.0253
Epoch 898/1000
81/81 [==============================] - 0s - loss: 0.0254 - val_loss: 0.0252
Epoch 899/1000
81/81 [==============================] - 0s - loss: 0.0252 - val_loss: 0.0252
Epoch 900/1000
81/81 [==============================] - 0s - loss: 0.0252 - val_loss: 0.0251
Epoch 901/1000
81/81 [==============================] - 0s - loss: 0.0252 - val_loss: 0.0251
Epoch 902/1000
81/81 [==============================] - 0s - loss: 0.0251 - val_loss: 0.0251
Epoch 903/1000
81/81 [==============================] - 0s - loss: 0.0252 - val_loss: 0.0250
Epoch 904/1000
81/81 [==============================] - 0s - loss: 0.0250 - val_loss: 0.0249
Epoch 905/1000
81/81 [==============================] - 0s - loss: 0.0250 - val_loss: 0.0249
Epoch 906/1000
81/81 [==============================] - 0s - loss: 0.0250 - val_loss: 0.0249
Epoch 907/1000
81/81 [==============================] - 0s - loss: 0.0249 - val_loss: 0.0248
Epoch 908/1000
81/81 [==============================] - 0s - loss: 0.0249 - val_loss: 0.0247
Epoch 909/1000
81/81 [==============================] - 0s - loss: 0.0248 - val_loss: 0.0246
Epoch 910/1000
81/81 [==============================] - 0s - loss: 0.0245 - val_loss: 0.0245
Epoch 911/1000
81/81 [==============================] - 0s - loss: 0.0246 - val_loss: 0.0245
Epoch 912/1000
81/81 [==============================] - 0s - loss: 0.0246 - val_loss: 0.0245
Epoch 913/1000
81/81 [==============================] - 0s - loss: 0.0244 - val_loss: 0.0243
Epoch 914/1000
81/81 [==============================] - 0s - loss: 0.0244 - val_loss: 0.0243
Epoch 915/1000
81/81 [==============================] - 0s - loss: 0.0244 - val_loss: 0.0243
Epoch 916/1000
81/81 [==============================] - 0s - loss: 0.0243 - val_loss: 0.0241
Epoch 917/1000
81/81 [==============================] - 0s - loss: 0.0241 - val_loss: 0.0241
Epoch 918/1000
81/81 [==============================] - 0s - loss: 0.0242 - val_loss: 0.0242
Epoch 919/1000
81/81 [==============================] - 0s - loss: 0.0242 - val_loss: 0.0241
Epoch 920/1000
81/81 [==============================] - 0s - loss: 0.0241 - val_loss: 0.0240
Epoch 921/1000
81/81 [==============================] - 0s - loss: 0.0241 - val_loss: 0.0241
Epoch 922/1000
81/81 [==============================] - 0s - loss: 0.0241 - val_loss: 0.0240
Epoch 923/1000
81/81 [==============================] - 0s - loss: 0.0241 - val_loss: 0.0238
Epoch 924/1000
81/81 [==============================] - 0s - loss: 0.0238 - val_loss: 0.0238
Epoch 925/1000
81/81 [==============================] - 0s - loss: 0.0239 - val_loss: 0.0239
Epoch 926/1000
81/81 [==============================] - 0s - loss: 0.0240 - val_loss: 0.0238
Epoch 927/1000
81/81 [==============================] - 0s - loss: 0.0238 - val_loss: 0.0236
Epoch 928/1000
81/81 [==============================] - 0s - loss: 0.0236 - val_loss: 0.0237
Epoch 929/1000
81/81 [==============================] - 0s - loss: 0.0238 - val_loss: 0.0237
Epoch 930/1000
81/81 [==============================] - 0s - loss: 0.0238 - val_loss: 0.0236
Epoch 931/1000
81/81 [==============================] - 0s - loss: 0.0235 - val_loss: 0.0233
Epoch 932/1000
81/81 [==============================] - 0s - loss: 0.0234 - val_loss: 0.0234
Epoch 933/1000
81/81 [==============================] - 0s - loss: 0.0235 - val_loss: 0.0234
Epoch 934/1000
81/81 [==============================] - 0s - loss: 0.0235 - val_loss: 0.0233
Epoch 935/1000
81/81 [==============================] - 0s - loss: 0.0233 - val_loss: 0.0232
Epoch 936/1000
81/81 [==============================] - 0s - loss: 0.0232 - val_loss: 0.0231
Epoch 937/1000
81/81 [==============================] - 0s - loss: 0.0231 - val_loss: 0.0230
Epoch 938/1000
81/81 [==============================] - 0s - loss: 0.0231 - val_loss: 0.0230
Epoch 939/1000
81/81 [==============================] - 0s - loss: 0.0231 - val_loss: 0.0230
Epoch 940/1000
81/81 [==============================] - 0s - loss: 0.0231 - val_loss: 0.0230
Epoch 941/1000
81/81 [==============================] - 0s - loss: 0.0230 - val_loss: 0.0230
Epoch 942/1000
81/81 [==============================] - 0s - loss: 0.0230 - val_loss: 0.0230
Epoch 943/1000
81/81 [==============================] - 0s - loss: 0.0231 - val_loss: 0.0229
Epoch 944/1000
81/81 [==============================] - 0s - loss: 0.0230 - val_loss: 0.0229
Epoch 945/1000
81/81 [==============================] - 0s - loss: 0.0228 - val_loss: 0.0228
Epoch 946/1000
81/81 [==============================] - 0s - loss: 0.0228 - val_loss: 0.0227
Epoch 947/1000
81/81 [==============================] - 0s - loss: 0.0227 - val_loss: 0.0227
Epoch 948/1000
81/81 [==============================] - 0s - loss: 0.0227 - val_loss: 0.0227
Epoch 949/1000
81/81 [==============================] - 0s - loss: 0.0228 - val_loss: 0.0227
Epoch 950/1000
81/81 [==============================] - 0s - loss: 0.0228 - val_loss: 0.0227
Epoch 951/1000
81/81 [==============================] - 0s - loss: 0.0226 - val_loss: 0.0225
Epoch 952/1000
81/81 [==============================] - 0s - loss: 0.0227 - val_loss: 0.0225
Epoch 953/1000
81/81 [==============================] - 0s - loss: 0.0226 - val_loss: 0.0224
Epoch 954/1000
81/81 [==============================] - 0s - loss: 0.0224 - val_loss: 0.0223
Epoch 955/1000
81/81 [==============================] - 0s - loss: 0.0224 - val_loss: 0.0223
Epoch 956/1000
81/81 [==============================] - 0s - loss: 0.0223 - val_loss: 0.0222
Epoch 957/1000
81/81 [==============================] - 0s - loss: 0.0222 - val_loss: 0.0221
Epoch 958/1000
81/81 [==============================] - 0s - loss: 0.0222 - val_loss: 0.0221
Epoch 959/1000
81/81 [==============================] - 0s - loss: 0.0221 - val_loss: 0.0220
Epoch 960/1000
81/81 [==============================] - 0s - loss: 0.0221 - val_loss: 0.0221
Epoch 961/1000
81/81 [==============================] - 0s - loss: 0.0222 - val_loss: 0.0221
Epoch 962/1000
81/81 [==============================] - 0s - loss: 0.0221 - val_loss: 0.0220
Epoch 963/1000
81/81 [==============================] - 0s - loss: 0.0220 - val_loss: 0.0219
Epoch 964/1000
81/81 [==============================] - 0s - loss: 0.0220 - val_loss: 0.0218
Epoch 965/1000
81/81 [==============================] - 0s - loss: 0.0218 - val_loss: 0.0216
Epoch 966/1000
81/81 [==============================] - 0s - loss: 0.0217 - val_loss: 0.0216
Epoch 967/1000
81/81 [==============================] - 0s - loss: 0.0216 - val_loss: 0.0216
Epoch 968/1000
81/81 [==============================] - 0s - loss: 0.0216 - val_loss: 0.0215
Epoch 969/1000
81/81 [==============================] - 0s - loss: 0.0216 - val_loss: 0.0214
Epoch 970/1000
81/81 [==============================] - 0s - loss: 0.0215 - val_loss: 0.0214
Epoch 971/1000
81/81 [==============================] - 0s - loss: 0.0216 - val_loss: 0.0214
Epoch 972/1000
81/81 [==============================] - 0s - loss: 0.0215 - val_loss: 0.0213
Epoch 973/1000
81/81 [==============================] - 0s - loss: 0.0213 - val_loss: 0.0213
Epoch 974/1000
81/81 [==============================] - 0s - loss: 0.0213 - val_loss: 0.0212
Epoch 975/1000
81/81 [==============================] - 0s - loss: 0.0213 - val_loss: 0.0211
Epoch 976/1000
81/81 [==============================] - 0s - loss: 0.0212 - val_loss: 0.0211
Epoch 977/1000
81/81 [==============================] - 0s - loss: 0.0211 - val_loss: 0.0210
Epoch 978/1000
81/81 [==============================] - 0s - loss: 0.0211 - val_loss: 0.0210
Epoch 979/1000
81/81 [==============================] - 0s - loss: 0.0211 - val_loss: 0.0210
Epoch 980/1000
81/81 [==============================] - 0s - loss: 0.0211 - val_loss: 0.0210
Epoch 981/1000
81/81 [==============================] - 0s - loss: 0.0210 - val_loss: 0.0209
Epoch 982/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0208
Epoch 983/1000
81/81 [==============================] - 0s - loss: 0.0210 - val_loss: 0.0209
Epoch 984/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0208
Epoch 985/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0207
Epoch 986/1000
81/81 [==============================] - 0s - loss: 0.0207 - val_loss: 0.0207
Epoch 987/1000
81/81 [==============================] - 0s - loss: 0.0207 - val_loss: 0.0209
Epoch 988/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0209
Epoch 989/1000
81/81 [==============================] - 0s - loss: 0.0208 - val_loss: 0.0206
Epoch 990/1000
81/81 [==============================] - 0s - loss: 0.0207 - val_loss: 0.0206
Epoch 991/1000
81/81 [==============================] - 0s - loss: 0.0207 - val_loss: 0.0207
Epoch 992/1000
81/81 [==============================] - 0s - loss: 0.0208 - val_loss: 0.0206
Epoch 993/1000
81/81 [==============================] - 0s - loss: 0.0205 - val_loss: 0.0205
Epoch 994/1000
81/81 [==============================] - 0s - loss: 0.0205 - val_loss: 0.0208
Epoch 995/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0209
Epoch 996/1000
81/81 [==============================] - 0s - loss: 0.0209 - val_loss: 0.0206
Epoch 997/1000
81/81 [==============================] - 0s - loss: 0.0206 - val_loss: 0.0203
Epoch 998/1000
81/81 [==============================] - 0s - loss: 0.0203 - val_loss: 0.0204
Epoch 999/1000
81/81 [==============================] - 0s - loss: 0.0204 - val_loss: 0.0205
Epoch 1000/1000
81/81 [==============================] - 0s - loss: 0.0206 - val_loss: 0.0203
      1   2   3   4   5   6   7   8   9
  1   1   2   3   4   5   6   7   8   9
  2   2   4   6   8  10  12  14  16  18
  3   3   6   1  12  15  18  21  24  27
  4   4   8  12  16  20  24  28  32  36
  5   5  10  15  28  17  30  35  40  45
  6   6  12  18  24  30  44  42  48  54
  7   7  14  21  28  35  42  49  56  63
  8   8  16  24  32  40  48  56  64  72
  9   9  18  27  36  45  54  63  72  81

サンプルコード

import numpy as np
import tensorflow as tf
from tensorflow.contrib.keras.python import keras
from sklearn import cross_validation
from tensorflow.contrib.keras.python.keras.models import Sequential
from tensorflow.contrib.keras.python.keras.layers import Dense, Dropout, Embedding, LSTM

def in_encode(i, j):
    k = j * 16 + i
    return np.array([k >> d & 1 for d in range(8)])

def out_encode(i, j):
    k = j * i
    return np.array([k >> d & 1 for d in range(7)])

def decode(p):
    f = 0
    if p[0] > 0.5:
        f += 1
    if p[1] > 0.5:
        f += 2
    if p[2] > 0.5:
        f += 4
    if p[3] > 0.5:
        f += 8
    if p[4] > 0.5:
        f += 16
    if p[5] > 0.5:
        f += 32
    if p[6] > 0.5:
        f += 64
    return f

trX = np.array([in_encode(i, j) for i in range(1, 10) for j in range(1, 10)])
trY = np.array([out_encode(i, j) for i in range(1, 10) for j in range(1, 10)])
model = Sequential()
model.add(Dense(40, activation = 'tanh', input_shape = (8, )))
model.add(Dense(40, activation = 'tanh'))
model.add(Dense(7, activation = 'linear'))
model.compile(loss = 'mean_squared_error', optimizer = 'adam')
history = model.fit(trX, trY, batch_size = 60, epochs = 1000, verbose = 1, validation_data = (trX, trY))
score = model.evaluate(trX, trY, verbose = 0)
p = '    '
j = 1
for i in range(1, 10):
    p += '%3d ' % (i * j)
p += '\n'
for j in range(1, 10):
    p += '%3d ' % (j)
    for i in range(1, 10):
        x = np.array([in_encode(i, j)])
        pred = model.predict([x])
        k = decode(pred[0])
        p += '%3d ' % (k)
    p += '\n'
print (p)

以上。