# numpy_neural_network

**Repository Path**: wangwanli666/numpy_neural_network

## Basic Information

- **Project Name**: numpy_neural_network
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2023-11-20
- **Last Updated**: 2023-11-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# numpy_neural_network

#### 介绍
该项目为使用numpy实现CNN的前向和反向传播

仓库地址：https://gitee.com/wangwanli666/numpy_neural_network

#### 项目结构
    activation:
        - relu.py
        - softmax.py
    data:
        - mnist
    layers:
        - convolutional.py
        - dense.py
        - dropout.py
        - flatten.py
        - pooling.py
    model:
        - sequential.py
    optimizers:
        - gradient_descent.py
    utils:
        - error.py
        - metrics.py
        - util.py
    base.py: 定义了layers和optimizer的基类

### 模型生成
    在model/sequential.py中，使用SequentialModel来生成模型，需要传入定义的网络结构和优化器，如下代码所示
    layers = [
        ConvLayer2D.initialize(filters=16, kernel_shape=(3, 3, 1), stride=1),
        ReluLayer(),
        MaxPoolLayer(pool_size=(2, 2), stride=2)
        FlattenLayer(),
        DenseLayer.initialize(units_prev=288, units_curr=64),
        ReluLayer(),
        DenseLayer.initialize(units_prev=64, units_curr=N_CLASSES),
        SoftmaxLayer()
    ]
    optimizer = GradientDescent(lr=LR)
    model = SequentialModel(
        layers=layers,
        optimizer=optimizer
    )

#### 实例
    demo.py: 使用mnist数据集，训练手写数字识别的CNN模型