def main(): encode = np.ones((10, 10)) for i in xrange(10): encode[i, i] = 0.9 print " start reading train data : ", time.time() traindata_path = os.path.join(datadir, "train.csv") traindata = LoadData.getCSV(traindata_path, np.float) h_matrix = traindata[:, 1:784] print time.time() np.save("h_matrix.npy", h_matrix) print " finish reading train data : ", time.time() print h_matrix.shape target = traindata[:, 0] target_s = [encode[i] for i in target] target = np.array(target_s) np.save("target.npy", target) print "start training my model : ", time.time() weights = ANN.BackPropagation(h_matrix / 255.0, target, 2, rate=0.3, iter_times=100) print "train finished : ", time.time() np.save("weights.npy", weights) print " start reading test data : ", time.time() testdata_path = os.path.join(datadir, "test.csv") testdata = LoadData.getCSV(testdata_path, np.float) test_h_matrix = testdata[:, 1:784] print "start predicting : ", time.time() result = ANN.get_predict(test_h_matrix, weights, 784, 2, 10) print "finish predicting : ", time.time() np.save("result.npy", result)
def testANNPerceptron(): init_weights = np.array([0.0, 0.0, 0.0]) data = LoadData.getCSV("xor.csv", np.float, firstline=False) target = data[:, data.shape[1] - 1] h_matrix = data[:, :data.shape[1] - 1] weights, residual = ANN.PerceptronRule( h_matrix, target, init_weights, rate=0.2, iter_times=100, perceptron=True) print weights print residual
def readTestData(): testdata_path = os.path.join(datadir, "test.csv") testdata = LoadData.getCSV(testdata_path, np.float) test_h_matrix = testdata[:, 1:784] np.save("testdata.npy", test_h_matrix)