def load_one_data(path):
data=[]
print('load_one_data')
load.load(data,path,'jpg',num_ide=77,num_exp=47,num_land=73)
print(data)
return data
def load_set_data():
data=[]
fwhs_path='/home/weiliu/fitting_dde/fitting_psp_f_l12_slt/fw'
lfw_path='/home/weiliu/fitting_dde/fitting_psp_f_l12_slt/lfw_image'
gtav_path='/home/weiliu/fitting_dde/fitting_psp_f_l12_slt/GTAV_image'
load.load(data,fwhs_path,'jpg',num_ide=77,num_exp=47,num_land=73)
load.load(data,lfw_path,'jpg',num_ide=77,num_exp=47,num_land=73)
load.load(data,gtav_path,'png',num_ide=77,num_exp=47,num_land=73)
# test_path='/home/weiliu/fitting_dde/4_psp_f_cal_test/data_me/test_only_three/'
# load.load(data,test_path,'jpg',num_ide=77,num_exp=47,num_land=73)
return data
_, Z = forward(X, Node, W, B, h)
L = len(Node)
arg = np.argmax(Z[L-1], axis=0)
return np.average(arg == t)
def noise_on_train(X):
num_rand = np.random.binomial(n=784*N, p=noise*0.05)
rand_mask_arg = np.random.choice(784*N, num_rand, replace=False)
rand_mask = np.zeros(784*N).astype(np.bool_)
rand_mask[rand_mask_arg] = True
rand_mask.shape = [784, N]
X[rand_mask] = np.random.rand(num_rand)
return X
if __name__ == "__main__":
dataset = load.load()
t_train = one_of_k(dataset["train_labels"]).T #shape (10, 60000)
t_test = dataset["test_labels"]
X_train = dataset["train_images"].T/255 #shape (784, 60000)
X_test = dataset["test_images"].T/255 #shape (784, 10000)
#Hyper Param
N = 1000 # number of minibatch
M = 2000 # maximum iterations (epoch)
e = 0.1 # learning rate
w_init_mean = 0
w_init_sigma = 0.1
Node = [784, 256, 128, 64, 10]
L = len(Node)