File size: 5,862 Bytes
7b5e67a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 |
import os
import json
import numpy as np
import pandas as pd
import matplotlib as mpl
import seaborn as sns
def main():
# hyperparameters
datasets = ["mnist", "fmnist", "cifar10"]
selected_epochs_dict = {"mnist":[5],"fmnist":[2,6,11], "cifar10":[3,9,18,41]}
k_neighbors = [10,15,20]
col = np.array(["dataset", "method", "type", "hue", "k", "period", "eval"])
df = pd.DataFrame({}, columns=col)
for k in k_neighbors: # k neighbors
for i in range(len(datasets)): # dataset
dataset = datasets[i]
data = np.array([])
selected_epochs = selected_epochs_dict[dataset]
# DeepDebugger smoothness
eval_path = "/home/xianglin/projects/DVI_data/resnet18_{}/Model/test_evaluation_hybrid.json".format(dataset)
with open(eval_path, "r") as f:
eval = json.load(f)
for epoch_id in range(len(selected_epochs)):
epoch = selected_epochs[epoch_id]
nn_train = round(eval["nn_train"][str(epoch)][str(k)], 3)
nn_test = round(eval["nn_test"][str(epoch)][str(k)], 3)
if len(data) == 0:
data = np.array([[dataset, "DeepDebugger", "Train", "DeepDebugger(Train)", "{}".format(k), "{}".format(str(epoch)), nn_train]])
else:
data = np.concatenate((data, np.array([[dataset, "DeepDebugger", "Train", "DeepDebugger(Train)", "{}".format(k), "{}".format(str(epoch)), nn_train]])), axis=0)
data = np.concatenate((data, np.array([[dataset, "DeepDebugger", "Test", "DeepDebugger(Test)", "{}".format(k), "{}".format(str(epoch)), nn_test]])), axis=0)
# DeepDebugger without smoothness
eval_path = "/home/xianglin/projects/DVI_data/resnet18_{}/Model/without_smoothness/test_evaluation_hybrid.json".format(dataset)
with open(eval_path, "r") as f:
eval = json.load(f)
for epoch_id in range(len(selected_epochs)):
epoch = selected_epochs[epoch_id]
nn_train = round(eval["nn_train"][str(epoch)][str(k)], 3)
nn_test = round(eval["nn_test"][str(epoch)][str(k)], 3)
data = np.concatenate((data, np.array([[dataset, "no_Smoothness", "Train", "-SS(Train)", "{}".format(k), "{}".format(str(epoch)), nn_train]])), axis=0)
data = np.concatenate((data, np.array([[dataset, "no_Smoothness", "Test", "-SS(Test)", "{}".format(k), "{}".format(str(epoch)), nn_test]])), axis=0)
# DeepDebugger without tl
eval_path = "/home/xianglin/projects/DVI_data/resnet18_{}/Model/without_tl/test_evaluation_hybrid.json".format(dataset)
with open(eval_path, "r") as f:
eval = json.load(f)
for epoch_id in range(len(selected_epochs)):
epoch = selected_epochs[epoch_id]
nn_train = round(eval["nn_train"][str(epoch)][str(k)], 3)
nn_test = round(eval["nn_test"][str(epoch)][str(k)], 3)
data = np.concatenate((data, np.array([[dataset, "no_TL", "Train", "-TL(Train)", "{}".format(k), "{}".format(str(epoch)), nn_train]])), axis=0)
data = np.concatenate((data, np.array([[dataset, "no_TL", "Test", "-TL(Test)", "{}".format(k), "{}".format(str(epoch)), nn_test]])), axis=0)
df_tmp = pd.DataFrame(data, columns=col)
df = df.append(df_tmp, ignore_index=True)
df[["period"]] = df[["period"]].astype(int)
df[["k"]] = df[["k"]].astype(int)
df[["eval"]] = df[["eval"]].astype(float)
df.to_excel("./plot_results/ablation_smoothness_nn.xlsx")
for k in k_neighbors:
df_tmp = df[df["k"] == k]
pal20c = sns.color_palette('tab20', 20)
sns.set_theme(style="whitegrid", palette=pal20c)
hue_dict = {
"-TL(Train)": pal20c[10],
"-SS(Train)": pal20c[12],
"DeepDebugger(Train)": pal20c[18],
"-TL(Test)": pal20c[11],
"-SS(Test)": pal20c[13],
"DeepDebugger(Test)": pal20c[19],
}
sns.palplot([hue_dict[i] for i in hue_dict.keys()])
axes = {'labelsize': 10,
'titlesize': 10,}
mpl.rc('axes', **axes)
mpl.rcParams['xtick.labelsize'] = 10
hue_list = ["-TL(Train)", "-TL(Test)", "-SS(Train)", "-SS(Test)", "DeepDebugger(Train)", "DeepDebugger(Test)"]
fg = sns.catplot(
x="period",
y="eval",
hue="hue",
hue_order=hue_list,
# order = [1, 2, 3, 4, 5],
# row="method",
col="dataset",
ci=0.001,
height=2.5, #2.65,
aspect=1.0,#3,
data=df_tmp,
kind="bar",
sharex=False,
palette=[hue_dict[i] for i in hue_list],
legend=True
)
sns.move_legend(fg, "lower center", bbox_to_anchor=(.42, 0.92), ncol=3, title=None, frameon=False)
mpl.pyplot.setp(fg._legend.get_texts(), fontsize='10')
axs = fg.axes[0]
# max_ = df_tmp["eval"].max()
# min_ = df["eval"].min()
# axs[0].set_ylim(0., max_*1.1)
axs[0].set_title("MNIST")
axs[1].set_title("FMNIST")
axs[2].set_title("CIFAR-10")
(fg.despine(bottom=False, right=False, left=False, top=False)
# .set_xticklabels(['Early', 'Mid', 'Late'])
.set_axis_labels("", "")
)
# fg.fig.suptitle("NN preserving property")
fg.savefig(
"./plot_results/ablation_smoothness_nn_{}.png".format(k),
dpi=300,
bbox_inches="tight",
pad_inches=0.0,
transparent=True,
)
if __name__ == "__main__":
main()
|