from params.parameters import Parameters tensorfuzz = Parameters() tensorfuzz.tf_num_mutations = 64 tensorfuzz.tf_sigma = 0.2 * (255 - 0) / (1 - (-1)) tensorfuzz.constraint = None
from params.parameters import Parameters LeNet4 = Parameters() LeNet4.tfc_threshold = 169 LeNet4.model_input_scale = [0, 1] LeNet4.skip_layers = [0, 5]
from params.parameters import Parameters LeNet5 = Parameters() LeNet5.tfc_threshold = 121 LeNet5.model_input_scale = [0, 1] LeNet5.skip_layers = [0, 5]
import numpy as np import itertools import src.image_transforms as image_transforms from params.parameters import Parameters deephunter = Parameters() deephunter.K = 64 deephunter.batch1 = 64 deephunter.batch2 = 16 deephunter.p_min = 0.01 deephunter.gamma = 5 deephunter.alpha = 0.1 deephunter.beta = 0.5 deephunter.TRY_NUM = 100 # translation = list(itertools.product([getattr(image_transforms,"image_translation")], [(10+10*k,10+10*k) for k in range(10)])) # scale = list(itertools.product([getattr(image_transforms, "image_scale")], [(1.5+0.5*k,1.5+0.5*k) for k in range(10)])) # shear = list(itertools.product([getattr(image_transforms, "image_shear")], [(-1.0+0.1*k,0) for k in range(10)])) # rotation = list(itertools.product([getattr(image_transforms, "image_rotation")], [3+3*k for k in range(10)])) # contrast = list(itertools.product([getattr(image_transforms, "image_contrast")], [1.2+0.2*k for k in range(10)])) # brightness = list(itertools.product([getattr(image_transforms, "image_brightness")], [10+10*k for k in range(10)])) # blur = list(itertools.product([getattr(image_transforms, "image_blur")], [k+1 for k in range(10)])) translation = list( itertools.product([getattr(image_transforms, "image_translation")], [(-5, -5), (-5, 0), (0, -5), (0, 0), (5, 0), (0, 5), (5, 5)])) rotation = list( itertools.product([getattr(image_transforms, "image_rotation")], [-15, -12, -9, -6, -3, 3, 6, 9, 12, 15]))
from params.parameters import Parameters kmn = Parameters() kmn.kmn_k = 10000
from params.parameters import Parameters LeNet1 = Parameters() LeNet1.tfc_threshold = 900 LeNet1.model_input_scale = [0, 1] LeNet1.skip_layers = [0, 5]
from params.parameters import Parameters CIFAR_CNN = Parameters() CIFAR_CNN.tfc_threshold = 9 CIFAR_CNN.model_input_scale = [0, 1] CIFAR_CNN.skip_layers = [0, 2, 5, 6, 8, 11, 12, 13, 15]
from params.parameters import Parameters neuron = Parameters()
import numpy as np import itertools from params.parameters import Parameters mcts = Parameters() def tc1(state): # limit the level/depth of root return state.level > 8 mcts.tc1 = tc1 def tc2(iterations): # limit the number of iterations on root return iterations > 25 mcts.tc2 = tc2 def tc3(state): original_input = state.original_input mutated_input = state.mutated_input alpha, beta = 0.1, 0.5 if(np.sum((original_input-mutated_input) != 0) < alpha * np.sum(original_input>0)): return not np.max(np.abs(mutated_input-original_input)) <= 255 else: return not np.max(np.abs(mutated_input-original_input)) <= beta*255 mcts.tc3 = tc3
from params.parameters import Parameters snac = Parameters()
from params.parameters import Parameters nbc = Parameters()
from params.parameters import Parameters cifar10 = Parameters() cifar10.input_shape = (1, 32, 32, 3) cifar10.input_lower_limit = 0 cifar10.input_upper_limit = 255
from params.parameters import Parameters mnist = Parameters() mnist.input_shape = (1, 28, 28, 1) mnist.input_lower_limit = 0 mnist.input_upper_limit = 255