def gen_gp_test_data():
""" Generates test data for the unit tests. """
# CNNs
cnns = generate_cnn_architectures()
np.random.shuffle(cnns)
X1_tr = cnns[:7]
Y1_tr = np.array([cnn_syn_func1(x) for x in X1_tr])
X1_te = cnns[7:]
Y1_te = np.array([cnn_syn_func1(x) for x in X1_te])
nn_type_1 = 'cnn'
# MLP regression
regmlps = generate_mlp_architectures('reg')
np.random.shuffle(regmlps)
X2_tr = regmlps[:5]
Y2_tr = np.array([mlp_syn_func1(x) for x in X2_tr])
X2_te = regmlps[5:]
Y2_te = np.array([mlp_syn_func1(x) for x in X2_te])
nn_type_2 = 'mlp-reg'
# MLP classification
classmlps = generate_mlp_architectures('class')
np.random.shuffle(classmlps)
X3_tr = classmlps[:5]
Y3_tr = np.array([syn_func1_common(x) for x in X3_tr])
X3_te = classmlps[5:]
Y3_te = np.array([syn_func1_common(x) for x in X3_te])
nn_type_3 = 'mlp-class'
return [(X1_tr, Y1_tr, X1_te, Y1_te, nn_type_1),
(X2_tr, Y2_tr, X2_te, Y2_te, nn_type_2),
(X3_tr, Y3_tr, X3_te, Y3_te, nn_type_3)]
def get_nn_opt_arguments():
""" Returns arguments for NN Optimisation. """
ret = Namespace()
ret.cnn_constraint_checker = CNNConstraintChecker(50, 5, 1e8, 0, 5, 5, 200, 1024, 8)
ret.mlp_constraint_checker = MLPConstraintChecker(50, 5, 1e8, 0, 5, 5, 200, 1024, 8)
ret.cnn_mutation_op = get_nn_modifier_from_args(ret.cnn_constraint_checker,
[0.5, 0.25, 0.125, 0.075, 0.05])
ret.mlp_mutation_op = get_nn_modifier_from_args(ret.mlp_constraint_checker,
[0.5, 0.25, 0.125, 0.075, 0.05])
# Create the initial pool
ret.cnn_init_pool = get_initial_cnn_pool()
ret.cnn_init_vals = [cnn_syn_func1(cnn) for cnn in ret.cnn_init_pool]
ret.mlp_init_pool = get_initial_mlp_pool('reg')
ret.mlp_init_vals = [mlp_syn_func1(mlp) for mlp in ret.mlp_init_pool]
# Create a domain
ret.nn_domain = NNDomain(None, None)
ret.cnn_domain = NNDomain('cnn', ret.cnn_constraint_checker)
ret.mlp_domain = NNDomain('mlp-reg', ret.mlp_constraint_checker)
# Return
return ret
def syn_cnn_1(x): """ Computes the Branin function. """ return cnn_syn_func1(x[0])