def test_reset_compute(self): network = RbfNetwork(2, 1, 1) total = 0 for i in xrange(0, len(network.long_term_memory)): total += network.long_term_memory[i] self.assertEquals(0, total) network.reset() total = 0 for i in xrange(0, len(network.long_term_memory)): total += network.long_term_memory[i] self.assertTrue(total > 1)
# Normalize iris species using one-of-n. # We could have used equilateral as well. For an example of equilateral, see the example_nm_iris example. norm.norm_col_one_of_n(iris_work, 4, classes, 0, 1) # Prepare training data. Separate into input and ideal. training = np.array(iris_work) training_input = training[:, 0:4] training_ideal = training[:, 4:7] # Create an RBF network. There are four inputs and two outputs. # There are also five RBF functions used internally. # You can experiment with different numbers of internal RBF functions. # However, the input and output must match the data set. network = RbfNetwork(4, 4, 3) network.reset() def score_funct(x): """ The score function for Iris anneal. @param x: @return: """ global best_score global input_data global output_data # Update the network's long term memory to the vector we need to score. network.copy_memory(x) # Loop over the training set and calculate the output for each. actual_output = []
# Normalize iris species using one-of-n. # We could have used equilateral as well. For an example of equilateral, see the example_nm_iris example. norm.norm_col_one_of_n(iris_work, 4, classes, 0, 1) # Prepare training data. Separate into input and ideal. training = np.array(iris_work) training_input = training[:, 0:4] training_ideal = training[:, 4:7] # Create an RBF network. There are four inputs and two outputs. # There are also five RBF functions used internally. # You can experiment with different numbers of internal RBF functions. # However, the input and output must match the data set. network = RbfNetwork(4, 4, 3) network.reset() def score_funct(x): """ The score function for Iris anneal. @param x: @return: """ global best_score global input_data global output_data # Update the network's long term memory to the vector we need to score. network.copy_memory(x) # Loop over the training set and calculate the output for each. actual_output = [] for input_data in training_input: