提交 20ad098b 编写于 作者: M mindspore-ci-bot 提交者: Gitee

!193 Update tutorial of differential privacy.

Merge pull request !193 from ZhidanLiu/master
...@@ -56,6 +56,9 @@ from mindarmour.diff_privacy import PrivacyMonitorFactory ...@@ -56,6 +56,9 @@ from mindarmour.diff_privacy import PrivacyMonitorFactory
from mindarmour.utils.logger import LogUtil from mindarmour.utils.logger import LogUtil
from lenet5_net import LeNet5 from lenet5_net import LeNet5
from lenet5_config import mnist_cfg as cfg from lenet5_config import mnist_cfg as cfg
LOGGER.set_level('INFO')
TAG = 'Lenet5_train'
``` ```
### 配置环境信息 ### 配置环境信息
...@@ -72,17 +75,17 @@ from lenet5_config import mnist_cfg as cfg ...@@ -72,17 +75,17 @@ from lenet5_config import mnist_cfg as cfg
- initial_noise_multiplier:差分隐私参数,高斯噪声的标准差等于initial_noise_multiplier乘以l2_norm_bound。 - initial_noise_multiplier:差分隐私参数,高斯噪声的标准差等于initial_noise_multiplier乘以l2_norm_bound。
```python ```python
parser = argparse.ArgumentParser(description='MindSpore Example') parser = argparse.ArgumentParser(description='MindSpore MNIST Example')
parser.add_argument('--device_target', type=str, default="Ascend", choices=['Ascend', 'GPU', 'CPU'], parser.add_argument('--device_target', type=str, default="Ascend", choices=['Ascend', 'GPU', 'CPU'],
help='device where the code will be implemented (default: Ascend)') help='device where the code will be implemented (default: Ascend)')
parser.add_argument('--data_path', type=str, default="./_unzip", parser.add_argument('--data_path', type=str, default="./MNIST_unzip",
help='path where the dataset is saved') help='path where the dataset is saved')
parser.add_argument('--dataset_sink_mode', type=bool, default=False, help='dataset_sink_mode is False or True') parser.add_argument('--dataset_sink_mode', type=bool, default=False, help='dataset_sink_mode is False or True')
parser.add_argument('--micro_batches', type=int, default=None, parser.add_argument('--micro_batches', type=int, default=32,
help='optional, if use differential privacy, need to set micro_batches') help='optional, if use differential privacy, need to set micro_batches')
parser.add_argument('--l2_norm_bound', type=float, default=1, parser.add_argument('--l2_norm_bound', type=float, default=1.0,
help='optional, if use differential privacy, need to set l2_norm_bound') help='optional, if use differential privacy, need to set l2_norm_bound')
parser.add_argument('--initial_noise_multiplier', type=float, default=0.1, parser.add_argument('--initial_noise_multiplier', type=float, default=1.5,
help='optional, if use differential privacy, need to set initial_noise_multiplier') help='optional, if use differential privacy, need to set initial_noise_multiplier')
args = parser.parse_args() args = parser.parse_args()
``` ```
...@@ -100,7 +103,7 @@ from lenet5_config import mnist_cfg as cfg ...@@ -100,7 +103,7 @@ from lenet5_config import mnist_cfg as cfg
加载数据集并处理成MindSpore数据格式。 加载数据集并处理成MindSpore数据格式。
```python ```python
def generate__dataset(data_path, batch_size=32, repeat_size=1, def generate_mnist_dataset(data_path, batch_size=32, repeat_size=1,
num_parallel_workers=1, sparse=True): num_parallel_workers=1, sparse=True):
""" """
create dataset for training or testing create dataset for training or testing
...@@ -167,7 +170,7 @@ def fc_with_initialize(input_channels, out_channels): ...@@ -167,7 +170,7 @@ def fc_with_initialize(input_channels, out_channels):
def weight_variable(): def weight_variable():
return TruncatedNormal(0.02) return TruncatedNormal(0.05)
class LeNet5(nn.Cell): class LeNet5(nn.Cell):
...@@ -201,7 +204,7 @@ class LeNet5(nn.Cell): ...@@ -201,7 +204,7 @@ class LeNet5(nn.Cell):
return x return x
``` ```
加载`LeNet`网络,定义损失函数、配置checkpoint、用上述定义的数据加载函数`generate__dataset`载入数据。 加载`LeNet`网络,定义损失函数、配置checkpoint、用上述定义的数据加载函数`generate_mnist_dataset`载入数据。
```python ```python
network = LeNet5() network = LeNet5()
...@@ -212,7 +215,7 @@ ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet", ...@@ -212,7 +215,7 @@ ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet",
directory='./trained_ckpt_file/', directory='./trained_ckpt_file/',
config=config_ck) config=config_ck)
ds_train = generate__dataset(os.path.join(args.data_path, "train"), ds_train = generate_mnist_dataset(os.path.join(args.data_path, "train"),
cfg.batch_size, cfg.batch_size,
cfg.epoch_size) cfg.epoch_size)
``` ```
...@@ -221,12 +224,15 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), ...@@ -221,12 +224,15 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"),
1. 配置差分隐私优化器的参数。 1. 配置差分隐私优化器的参数。
- 判断micro_batches和batch_size参数是否符合要求。
- 实例化差分隐私工厂类。 - 实例化差分隐私工厂类。
- 设置差分隐私的噪声机制,目前支持固定标准差的高斯噪声机制:'Gaussian'和自适应调整标准差的自适应高斯噪声机制:'AdaGaussian'。 - 设置差分隐私的噪声机制,目前支持固定标准差的高斯噪声机制:'Gaussian'和自适应调整标准差的自适应高斯噪声机制:'AdaGaussian'。
- 设置优化器类型,目前支持'SGD'和'Momentum'。 - 设置优化器类型,目前支持'SGD'和'Momentum'。
- 设置差分隐私预算监测器RDP,用于观测每个step中的差分隐私预算$\epsilon$的变化。 - 设置差分隐私预算监测器RDP,用于观测每个step中的差分隐私预算$\epsilon$的变化。
```python ```python
if args.micro_batches and cfg.batch_size % args.micro_batches != 0:
raise ValueError("Number of micro_batches should divide evenly batch_size")
gaussian_mech = DPOptimizerClassFactory(args.micro_batches) gaussian_mech = DPOptimizerClassFactory(args.micro_batches)
gaussian_mech.set_mechanisms('Gaussian', gaussian_mech.set_mechanisms('Gaussian',
norm_bound=args.l2_norm_bound, norm_bound=args.l2_norm_bound,
...@@ -236,9 +242,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), ...@@ -236,9 +242,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"),
momentum=cfg.momentum) momentum=cfg.momentum)
rdp_monitor = PrivacyMonitorFactory.create('rdp', rdp_monitor = PrivacyMonitorFactory.create('rdp',
num_samples=60000, num_samples=60000,
batch_size=16, batch_size=cfg.batch_size,
initial_noise_multiplier=5, initial_noise_multiplier=args.initial_noise_multiplier*
target_delta=0.5, args.l2_norm_bound,
per_print_times=10) per_print_times=10)
``` ```
...@@ -262,13 +268,13 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), ...@@ -262,13 +268,13 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"),
dataset_sink_mode=args.dataset_sink_mode) dataset_sink_mode=args.dataset_sink_mode)
LOGGER.info(TAG, "============== Starting Testing ==============") LOGGER.info(TAG, "============== Starting Testing ==============")
ckpt_file_name = 'trained_ckpt_file/checkpoint_lenet-10_1875.ckpt' ckpt_file_name = 'trained_ckpt_file/checkpoint_lenet-10_234.ckpt'
param_dict = load_checkpoint(ckpt_file_name) param_dict = load_checkpoint(ckpt_file_name)
load_param_into_net(network, param_dict) load_param_into_net(network, param_dict)
ds_eval = generate__dataset(os.path.join(args.data_path, 'test'), batch_size=cfg.batch_size) ds_eval = generate_mnist_dataset(os.path.join(args.data_path, 'test'), batch_size=cfg.batch_size)
acc = model.eval(ds_eval, dataset_sink_mode=False) acc = model.eval(ds_eval, dataset_sink_mode=False)
LOGGER.info(TAG, "============== Accuracy: %s ==============", acc) LOGGER.info(TAG, "============== Accuracy: %s ==============", acc)
``` ```
4. 运行命令。 4. 运行命令。
...@@ -290,11 +296,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), ...@@ -290,11 +296,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"),
... ...
============== Starting Testing ============== ============== Starting Testing ==============
... ...
============== Accuracy: 0.9635 ============== ============== Accuracy: 0.9091 ==============
``` ```
![dp_res](images/dp_res.png)
### 引用 ### 引用
[1] C. Dwork and J. Lei. Differential privacy and robust statistics. In STOC, pages 371–380. ACM, 2009. [1] C. Dwork and J. Lei. Differential privacy and robust statistics. In STOC, pages 371–380. ACM, 2009.
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册