This repository contains code for "Predicting Multi-Antenna Frequency-Selective Channels via Meta-Learned Linear Filters based on Long-Short Term Channel Decomposition" - Sangwoo Park and Osvaldo Simeone.

Dependencies

This program is written in python 3.8 and uses PyTorch 1.8.1.

Essential codes

• Naive and LSTD transfer-learning can be found at funcs/transfer_linear_filter.py.
• Naive meta-learning can be found at funcs/meta_linear_filter.py.
• LSTD meta-learning can be found at funcs/meta_lstd_linear_filter.py.
• Main file can be found at main_offline.py. Detailed usage can be found below.
• Channel dataset generation can be found in channel_gen folder.

How to run the codes

Prerequisites (data generation)

• Run channel_gen/5G_standard_SCM/main_custom.m to generate 5G standard SCM channel data (default: multi-antenna frequency-selective channel)

Train and Test

• For conventional learning (naive), execute runs/conven_naive.sh
• For conventional learning (LSTD), execute runs/conven_LSTD.sh
• For transfer learning (naive), execute runs/transfer_naive.sh
• For transfer learning (LSTD), execute runs/transfer_LSTD.sh
• For meta-learning (naive), execute runs/meta_naive.sh
• For meta-learning (LSTD), execute runs/meta_LSTD.sh