def load_data_sets(self):
if self.task == 'spam_enron':
self.data_sets = load_spam(ex_to_leave_out=self.ex_to_leave_out, num_examples=self.num_examples)
elif self.task == 'small_mnist':
self.data_sets = load_small_mnist(self.data_dir)
elif self.task == 'mnist':
self.data_sets = load_small_mnist('data')
elif self.task == 'heart_disease':
self.data_sets = load_heart_disease(ex_to_leave_out=self.ex_to_leave_out, num_examples=self.num_examples)
elif self.task == 'income':
self.data_sets = load_income(ex_to_leave_out=self.ex_to_leave_out, num_examples=self.num_examples)
if 'mnist' in self.task:
self.input_side = 28
self.input_channels = 1
self.input_dim = self.input_side * self.input_side * self.input_channels
else:
self.input_dim = self.data_sets.train.x.shape[1]
# import scipy
# import sklearn
import random
import influence.experiments as experiments
from influence.all_CNN_c import All_CNN_C
from load_mnist import load_small_mnist, load_mnist
import tensorflow as tf
# np.random.seed(42)
data_sets = load_small_mnist('data')
num_classes = 10
input_side = 28
input_channels = 1
input_dim = input_side * input_side * input_channels
weight_decay = 0.001
batch_size = 500
initial_learning_rate = 0.0001
decay_epochs = [10000, 20000]
hidden1_units = 8
hidden2_units = 8
hidden3_units = 8
conv_patch_size = 3
keep_probs = [1.0, 1.0]