Convolutional neural network based on numpy, modular design guarantees easy implementation of the model, which is suitable for the introduction of junior researchers in deep learning.
A PyTorch example is also included.
Realized Network Model（Located on the pynet/models）：
- 2-Layer Neural Network
- 3-Layer Neural Network
- LeNet-5
- AlexNet
- NIN
Realized Network Layer（Located on the pynet/nn）：
- Convolution Layer (Conv2d)
- Fully-Connection Layer (FC)
- Max-Pooling layer (MaxPool)
- ReLU Layer (ReLU)
- Random Dropout Layer (Dropout/Dropout2d)
- Softmax
- Cross Entropy Loss
- Gloabl Average Pool (GAP)
.
├── examples # pynet使用示例
│ ├── 2_nn_xor.py
│ ├── 3_nn_cifar10.py
│ ├── 3_nn_iris.py
│ ├── 3_nn_orl.py
│ ├── lenet5_mnist.py
│ ├── nin_cifar10.py
│ └── nin_cifar10_pytorch.py
├── plt # 绘图相关（待调整）
│ ├── anneal_plt.py
│ ├── lenet5_plt.py
│ └── plt.py
├── pynet # PyNet库
│ ├── __init__.py
│ ├── models # 模型定义
│ ├── nn # 层定义
│ └── vision # 数据操作
├── pytorch # PyTorch使用示例
│ ├── examples
│ ├── models # 模型定义
│ └── vision # 数据操作
We use SemVer for versioning. For the versions available, see the tags on this repository.
- zhujian - Initial work - zjZSTU
This project is licensed under the Apache License v2.0 - see the LICENSE file for details