我们从Python开源项目中，提取了以下8个代码示例，用于说明如何使用numpy.random.random_sample()。
def corrupt_image(img, MAR_prob=0, min_rects=0, max_rects=0, min_width=0, max_width=0): new_img = img.copy() mask = np.zeros(img.shape[0:2], dtype=np.bool) if MAR_prob > 0: mask[(random_sample(mask.shape) < MAR_prob)] = True if max_rects > 0 and max_width > 0: h, w = mask.shape num_rects = random_integers(min_rects, max_rects) for i in range(num_rects): px1 = random_integers(0, w - min(max(min_width, 1), w)) py1 = random_integers(0, h - min(max(min_width, 1), h)) px2 = px1 + (min_width - 1) + random_integers(0, max(min(w - px1 - min_width, max_width - min_width), 0)); py2 = py1 + (min_width - 1) + random_integers(0, max(min(h - py1 - min_width, max_width - min_width), 0)); if px1 <= px2 and py1 <= py2: mask[py1:py2, px1:px2] = True else: # One of the sides has length 0, so we should remove any pixels4 pass if len(new_img.shape) == 2: new_img[mask] = 0 else: new_img[mask,:] = 0 return (new_img, 1.0 * mask) # Process command line inputs
def sample_h_given_v(self, v): """ Given visible unit values, sample hidden unit values. Note: implemented in numpy for efficiency. Do not use in computation graph. """ mean_h = sigmoid(np.dot(v, self.params['W'].eval()) + self.params['bhid'].eval()) rnds = random_sample(mean_h.shape) return (mean_h > rnds).astype(np.float32)
def sample_v_given_h(self, h): """ Given hidden unit values, sample visible unit values. Note: implemented in numpy for efficiency. Do not use in computation graph. """ mean_v = sigmoid(np.dot(h, self.params['W'].eval().T) + self.params['bvis'].eval()) rnds = random_sample(mean_v.shape) return (mean_v > rnds).astype(np.float32)
def corrupt_image(img, MAR_prob=0, min_rects=0, max_rects=0, min_width=0, max_width=0, apply_to_all_channels=False): def generate_channel_mask(): mask = np.zeros(img.shape[0:2], dtype=np.bool) if MAR_prob > 0: mask[(random_sample(mask.shape) < MAR_prob)] = True if max_rects > 0 and max_width > 0: h, w = mask.shape num_rects = random_integers(min_rects, max_rects) for i in range(num_rects): px1 = random_integers(0, w - min(max(min_width, 1), w)) py1 = random_integers(0, h - min(max(min_width, 1), h)) px2 = px1 + min_width + random_integers(0, max(min(w - px1 - min_width, max_width - min_width), 0)); py2 = py1 + min_width + random_integers(0, max(min(h - py1 - min_width, max_width - min_width), 0)); if px1 <= px2 and py1 <= py2: mask[py1:py2, px1:px2] = True else: # One of the sides has length 0, so we should remove any pixels4 pass return mask new_img = img.copy() channels = 1 if len(new_img.shape) == 2 else new_img.shape[-1] global_mask = np.zeros(img.shape, dtype=np.bool) if channels == 1 or apply_to_all_channels: mask = generate_channel_mask() if channels == 1: global_mask[:, :] = mask else: for i in xrange(channels): global_mask[:, :, i] = mask else: global_mask = np.zeros(img.shape, dtype=np.bool) for i in xrange(channels): global_mask[:,:,i] = generate_channel_mask() new_img[global_mask] = 0 return (new_img, 1.0 * global_mask) # Process command line inputs
def simulate_df(n_rows, n_cols, n_categories=None, index_prefix='Index ', column_prefix='Column ', random_seed=RANDOM_SEED): """ Simulate DataFrame. Arguments: n_rows (int): number of rows n_cols (int): number of columns n_categories (None | int): None (for continuous) | int (for categorical) index_prefix (str): prefix for index column_prefix (str): prefix for column random_seed (int | array): Returns: DataFrame: (n_rows, n_cols) """ seed(random_seed) if n_categories: df = DataFrame(randint(0, n_categories, (n_rows, n_cols))) else: df = DataFrame(random_sample((n_rows, n_cols))) df.index = ['{}{}'.format(index_prefix, i) for i in range(n_rows)] df.columns = ['{}{}'.format(column_prefix, i) for i in range(n_cols)] return df
def simulate_series(size, n_categories=None, name='Simulated Series', index_prefix='Index ', random_seed=RANDOM_SEED): """ Simulate a Series. Arguments: size (int): size n_categories (None | int): None for continuous and int for categorical index_prefix (str): index prefix random_seed (int | array): Returns; Series: (size) """ seed(random_seed) if n_categories: series = Series(randint(low=0, high=n_categories - 1, size=size)) else: series = Series(random_sample(size)) series.name = name series.index = ['{}{}'.format(index_prefix, i) for i in range(size)] return series
def wolfe_line_search(self, fun, search_dir, curr_value, exp_decrease) : """ see Numerical Optimization, Nocedal and Wright, Algorithm 3.5, p. 60 """ f = lambda t : fun(self.after_step(t * search_dir))[0] fp = lambda t : self.scal_L2(fun(self.after_step(t * search_dir))[1], search_dir).Q0 exit_code = 0 # Default : everything is all right # Code to uncomment to check that fp is the true derivative of f======== h = 1e-8 for i in range(5) : t = random_sample() update_th = fp(t) update_emp = (f(t+h) - f(t-h)) / (2*h) print('') print('search dir : ', search_dir.to_array()) print('Checking the function passed to the Wolfe line search, t = ', t) print('Empirical derivative : ', update_emp) print('Theoretical derivative : ', update_th) #======================================================================= print("Exp decrease : ", exp_decrease) (a, _, _, _, _, _) = line_search(f, fp, 0, 1, exp_decrease, curr_value, c2 = 0.95) if a == None : print('Error during the wolfe line search') a = 0 exit_code = 1 # Exit_code = 1 : break ! step = a * search_dir new_state = self.after_step(step) self.set_state(new_state) #self.current_cost_grad = (C,grad) #self.is_current_cost_computed = True self.is_current_point_computed = False return (step, exit_code)
def setup_and_init_weights(nn_structure): W = {} b = {} for l in range(1, len(nn_structure)): W[l] = r.random_sample((nn_structure[l], nn_structure[l-1])) b[l] = r.random_sample((nn_structure[l],)) return W, b
