This is a homework of Deep Learning course.
In this homework, I build a simple fully connected(fc) Deep learning network. There is only one fc layer, one ReLU layer and one softmax layer. But it conatins the main idea of training the network - using backpropagation to update the weight and bais of the network. Then make the next epoch a higher accuracy theoretically.
To run the code, you need download all the files and put them into one folder. Extracting the zip file before you run the hw2.py by your favorite IDE.
中文解释：
此文件夹是CS519 深度学习课程的第二次作业。在这个作业中，我建立了一个一层全联接神经网络，一层ReLU层，最后还有一层Softmax输出层。虽然是个很简单的神经网络，但其中用到了最重要的反向传播算法，来提高下一循环的训练精确度。神经网络中每一层的功能都是手动编写，和用库文件相比，能更好的理解神经网络的运行原理。
使用指南：
- 下载所有文件至同一文件夹。
- 解压zip文件
- 运行hw2.py文件，并按提示操作。