This repository is used to record the project for the AIZU University course: Neural Networks I: Fundamental Theory and Applications.

Run the following code to use this repository.

git clone https://github.com/Jiaheliu137/neural_network
cd neural_network
pip install -r requirements.txt

1. project2(4-bit parity check):Report_project2_CN.md, Report_project2_EN.md, Report_project2_CN.pdf, Report_project2_EN.pdf

   cd project2

   1. Draw a graph

      python project2.py

   2. Draw a graph matrix

      python draw_plot.py

2. project3(Hopfield Network):Report_project3_CN.md, Report_project3_EN.md, Report_project3_CN.pdf, Report_project3_EN.pdf

   cd project3

   • Visualizing images of raw data patterns.

     python draw_pattern.py

   • Adding noise to raw data patterns.

     python noise.py -nl 0.15

     • $nl \in [0, 1]$

   • Recall noised pattern and visulize

     python project3.py -nl 15

     • $nl \in [0, 1]$ and $nl$ shuld be integer

3. project4(WTA network, Iris dataset):Report_project4_CN.md, Report_project4_EN.md, Report_project4_CN.pdf, Report_project4_EN.pdf

   cd project4
   python project4.py -v 3 -c 6

   • -v | --visualize : The dimensionality of visualization, $v \in {2, 3}$
   • -c | --cluster : Number of clusters, $c \geq 1$

4. project5(SOFM, Iris dataset):Report_project5_CN.md, Report_project5_EN.md, Report_project5_CN.pdf, Report_project5_EN.pdf

   • Use WTA to generate the cluster file

     python output_cluster.py [-c <num>]

   • Use SOFM to compressing data dimensions and visualize

     python project5.py [-c <num>]

The SOFM (Self-Organizing Feature Map) algorithm has multiple nested loops and requires plotting, which makes its execution speed relatively slow

eg.

cd project5
python output_cluster.py -c 3
python project5.py -c 3

• -c | --cluster : Number of clusters,the value is recommended $2 \leq c \leq 6$

5. Final_Presentation

​