AI Lab for Wireless Communication

This repository contains the projects and assignments for the course AI Lab for Wireless Communication. This course is designed to explore the applications of artificial intelligence and machine learning in wireless communication systems.

Projects

The following is a list of projects included in this repository:

Module 1

• Introduction to AI Algorithms for Channel Decoding

  2023-02-13

  Introduction to the lab, its tools and programming environment.

• Traditional Channel Decoding Algorithms

  2023-02-20

  Exploring two common decoding algorithms: syndrome decoding and maximum likelihood decoding. Syndrome decoding is a method for error correction that involves calculating the syndrome of the received codeword and using it to correct any errors. Maximum likelihood decoding is a more advanced decoding algorithm that uses statistical inference to determine the most likely transmitted codeword.

• Channel Decoding as Classification: Support Vector Machine

  2023-03-06

  Support Vector Machines (SVMs) are a type of machine learning algorithm that can be used for classification. In the context of channel decoding, the goal is to classify the received signal into one of the $2^4 = 16$ possible information messages. One approach is to use the one-versus-one method, which involves training the SVM ${{16}\choose{2}}$ times on each pair of possible classes.

• Channel Decoding as Classification: Deep Learning

  2023-03-13

  In this project, we will use TensorFlow, to explore how deep neural networks can be used for channel decoding. We will treat channel decoding as a classification problem, where the goal is to map the received signal to the most likely transmitted message. We will compare the performance of our deep learning-based approach with previous decoding algorithms and machine learning approaches.

• Final Project on AI for Channel Decoding

  2023-03-20

  For the final project of module 1, will apply the deep learning method on channel decoding as previous course. What's different is, we'll need to generate the message, encode the message with 15, 11 hamming code, modulate and assign noise on our own. This will require fully understanding of each course in module 1. The target will be to adjust the parameters till the performance of deep learning is close to maximum likelihood method at SNR=6.

Module 2

• Introduction to Deep Learning

  2023-03-27

  Establish a neural network for handwriting recognition of mnist dataset. The network includes DNN.

• Deep Learning and Convolutional Neural Network for Spot Localization

  2023-04-10

  Establish a neural network for indoor spot localization of the user. The network includes CNN, DNN, and other settings for optimization such as maxpooling, batch normalization, dropout.

Requirements

To run the code included in this repository, the following software and tools are required:

• Python 3.x
• Jupyter Notebook
• NumPy
• Matplotlib
• scikit-learn