File size: 4,743 Bytes
b92918a |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 |
import os
import sys
import itertools
import numpy as np
import tensorflow as tf
from sklearn.model_selection import train_test_split
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from src.model import MalConv
from src.utils import preprocess_dataset
def hyperparameter_search(csv_path,
param_grid=None,
max_length=2**20,
epochs=5,
validation_split=0.2):
"""
๊ทธ๋ฆฌ๋ ์์น๋ฅผ ํตํ ํ์ดํผํ๋ผ๋ฏธํฐ ์ต์ ํ
Args:
csv_path: ํ๋ จ ๋ฐ์ดํฐ CSV ๊ฒฝ๋ก
param_grid: ํ์ดํผํ๋ผ๋ฏธํฐ ๊ทธ๋ฆฌ๋
max_length: ์ต๋ ์
๋ ฅ ๊ธธ์ด
epochs: ํ๋ จ ์ํฌํฌ ์
validation_split: ๊ฒ์ฆ ๋ฐ์ดํฐ ๋น์จ
"""
if param_grid is None:
param_grid = {
'embedding_size': [8, 16],
'num_filters': [64, 128],
'fc_size': [64, 128],
'learning_rate': [0.001, 0.0001]
}
print("๋ฐ์ดํฐ ๋ก๋ฉ ์ค...")
X, y = preprocess_dataset(csv_path, max_length)
X_train, X_val, y_train, y_val = train_test_split(
X, y, test_size=validation_split, random_state=42, stratify=y
)
# ๋ชจ๋ ํ์ดํผํ๋ผ๋ฏธํฐ ์กฐํฉ ์์ฑ
param_names = list(param_grid.keys())
param_values = list(param_grid.values())
param_combinations = list(itertools.product(*param_values))
best_score = 0
best_params = None
results = []
print(f"์ด {len(param_combinations)}๊ฐ์ ์กฐํฉ์ ํ
์คํธํฉ๋๋ค.")
for i, params in enumerate(param_combinations):
param_dict = dict(zip(param_names, params))
print(f"\n[{i+1}/{len(param_combinations)}] ํ
์คํธ ์ค: {param_dict}")
try:
# ๋ชจ๋ธ ์์ฑ
model = MalConv(
max_input_length=max_length,
embedding_size=param_dict['embedding_size'],
num_filters=param_dict['num_filters'],
fc_size=param_dict['fc_size']
)
# ์ปดํ์ผ
model.compile(
optimizer=tf.keras.optimizers.Adam(
learning_rate=param_dict['learning_rate']
),
loss='binary_crossentropy',
metrics=['accuracy']
)
# ๋๋ฏธ ์
๋ ฅ์ผ๋ก ๋ชจ๋ธ ๋น๋
dummy_input = np.zeros((1, max_length), dtype=np.uint8)
_ = model(dummy_input)
# ํ๋ จ
history = model.fit(
X_train, y_train,
batch_size=16,
epochs=epochs,
validation_data=(X_val, y_val),
verbose=0
)
# ํ๊ฐ
val_loss, val_acc = model.evaluate(X_val, y_val, verbose=0)
result = {
'params': param_dict,
'val_accuracy': val_acc,
'val_loss': val_loss
}
results.append(result)
print(f"๊ฒ์ฆ ์ ํ๋: {val_acc:.4f}")
# ์ต๊ณ ์ฑ๋ฅ ์
๋ฐ์ดํธ
if val_acc > best_score:
best_score = val_acc
best_params = param_dict
print(f"์๋ก์ด ์ต๊ณ ์ฑ๋ฅ! ์ ํ๋: {best_score:.4f}")
except Exception as e:
print(f"์๋ฌ ๋ฐ์: {e}")
continue
print("\n" + "="*50)
print("ํ์ดํผํ๋ผ๋ฏธํฐ ํ๋ ์๋ฃ")
print("="*50)
print(f"์ต๊ณ ์ฑ๋ฅ: {best_score:.4f}")
print(f"์ต์ ํ์ดํผํ๋ผ๋ฏธํฐ: {best_params}")
# ๊ฒฐ๊ณผ ์ ๋ ฌ
results.sort(key=lambda x: x['val_accuracy'], reverse=True)
print("\n์์ 5๊ฐ ๊ฒฐ๊ณผ:")
for i, result in enumerate(results[:5]):
print(f"{i+1}. ์ ํ๋: {result['val_accuracy']:.4f}, "
f"ํ๋ผ๋ฏธํฐ: {result['params']}")
return best_params, results
def main():
csv_path = "Input/sample_data.csv" # ์ค์ ๋ฐ์ดํฐ ๊ฒฝ๋ก๋ก ๋ณ๊ฒฝ
# ์ปค์คํ
ํ์ดํผํ๋ผ๋ฏธํฐ ๊ทธ๋ฆฌ๋
param_grid = {
'embedding_size': [8, 16],
'num_filters': [64, 128],
'fc_size': [64, 128],
'learning_rate': [0.001, 0.0001]
}
best_params, results = hyperparameter_search(
csv_path=csv_path,
param_grid=param_grid,
epochs=3 # ๋น ๋ฅธ ํ
์คํธ๋ฅผ ์ํด ์ํฌํฌ ์ ๊ฐ์
)
print(f"\n์ต์ ํ์ดํผํ๋ผ๋ฏธํฐ๋ก ๋ชจ๋ธ์ ๋ค์ ํ๋ จํ์ธ์:")
print(f"python src/train.py {csv_path} --epochs 10")
if __name__ == "__main__":
main()
|