import numpy as np import matplotlib.pyplot as plt from PreprocessData import load37 from Equations import * import time # load train and test data: x, t = load37(version="train") x_test, t_test = load37(version="test") # store dimensions of data: N = np.shape(x)[0] d = np.shape(x)[1] w = np.random.randn(1, d) direction = np.random.randn(d) # set parameters: decay = 0.1 epochs = 250 losses = [] xaxis = [] # Start time: start = time.time() epoch = 0 step_size = 10 direction = 5 while np.sum(abs(step_size * direction)) > 0.3: if epoch % 10 == 0: print "Epoch: ", epoch
import numpy as np import matplotlib.pyplot as plt from PreprocessData import load37 from Equations import * import time # avoid overflow warnings np.seterr(all="ignore") # load train and test data: x_training, t_training = load37(version="train") x_test, t_test = load37(version="test") lb = 80 ub = 100 x = x_training[:lb] t = t_training[:lb] x_val = x_training[lb + 1:ub] t_val = t_training[lb + 1:ub] # store dimensions of data: N = np.shape(x)[0] d = np.shape(x)[1] # set parameters: decay = 0 epochs = 10000 eta = 0.1 alpha = 0.9 batch_size = 0.01 * ub
import matplotlib.pyplot as plt from Assignment2.Equations_mlp import * from PreprocessData import load37 # avoid overflow warnings np.seterr(all="ignore") # load data (N=12396L) x37_training, t37_training = load37() # lb = 499 # ub = 599 lb = 9999 ub = np.shape(x37_training)[0] - 1 x37_train = x37_training[:lb] t37_train = t37_training[:lb] x37_val = x37_training[lb+1:ub] t37_val = t37_training[lb+1:ub] # hyper parameters n_hidden = 30 n_output = 10 n_epochs = 10 eta = 0.0001 # initialize weights W1 = initialize_weights(x37_train.shape[1], n_hidden) W2 = initialize_weights(n_hidden, n_output)