Channel Prediction via Meta-Learning and EP

This repository contains code for "Predicting Flat-Fading Channels via Meta-Learned Closed-Form Linear Filters and Equilibrium Propagation" - Sangwoo Park and Osvaldo Simeone.

Dependencies

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

Essential codes

• Closed-form meta-learning for linear filters can be found at funcs/Ridge_meta.py.
• Gradient-based meta-learning for linear filters via Equilibrium Propagation (EP) can be found at funcs/Ridge_meta_EP.py.
• Offline meta-learning scenario can be found at main_offline.py. Detailed usage can be found below.
• Online meta-learning scenario can be found at main_online.py. Detailed usage can be found below.
• Channel dataset generation can be found at channel_gen folder.

How to run the codes

Prerequisites (data generation)

• For Random Doppler Frequency (Offline) (Fig. 2), run channel_gen/Jakes_Rounded/jakes_multi_w.m,rounded_multi_w.m and place the resulting .mat files into ../generated_channels/.
• For Standard Channel Model (Offline) (Fig. 3), run channel_gen/5G_standard_CDL/main_custom.m and place the resulting 3gpp_meta_training_offline.mat file into ../generated_channels/.
• For Gradient-Based Meta-Learning (Onlline) (Fig. 4), run channel_gen/5G_standard_CDL/main_custom.m and place the resulting .mat files into ../generated_channels/online_dataset/.

1) Random Doppler Frequency (Offline) (Fig. 2)

• For genie-aided performance, execute runs/fig_234/offline_jakes_rounded_fig_2/genie_aided.sh

• For conventional learning, execute runs/fig_234/offline_jakes_rounded_fig_2/conven.sh

• For joint learning, execute runs/fig_234/offline_jakes_rounded_fig_2/joint.sh

• For meta-learning, execute runs/fig_234/offline_jakes_rounded_fig_2/meta.sh

2) Standard Channel Model (Offline) (Fig. 3)

• For conventional learning, execute runs/fig_234/offline_standard_fig_3/conven.sh

• For joint learning, execute runs/fig_234/offline_standard_fig_3/joint.sh

• For meta-learning, execute runs/fig_234/offline_standard_fig_3/meta.sh

3) Gradient-Based Meta-Learning (Onlline) (Fig. 4)

• For EP-based online meta-learning, execute runs/fig_234/online_fig_4/online_meta_EP.sh

• For offline joint learning, execute runs/fig_234/online_fig_4/offline_joint.sh

• For offline meta-learning, execute runs/fig_234/online_fig_4/offline_meta.sh