ITE Technical Group Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2024-02-16 10:50
Adapter-Based Fine-Tuning for Multi-Task Learning Based CSI Feedback in FDD Massive MIMO Systems
Mayuko Inoue, Tomoaki Ohtsuki (Keio Univ)
Abstract (in Japanese) (See Japanese page) 
(in English) A multi-task learning-based Channel State Information (CSI) feedback has been proposed to obtain the CSI of the downlink communication channel at a Base Station (BS) in massive multiple-input multiple-output (MIMO) frequency-division-duplex (FDD) system. After pre-training the encoder and the decoder with an equal mix of CSI datasets from multiple channel environments, the decoder is fine-tuned with a small amount of CSI datasets from each channel environment. In this way, CSI feedback for multiple-channel environments can be realized with only one type of encoder in the User Equipment (UE). On the other hand, the BS needs to have enough decoders for each channel environment. We also believe that new expressions currently learned by fine-tuning in each layer of decoders can be learned more effectively by multiple layers. Based on these considerations, we propose adapter-based fine-tuning. We consider three types of Adapters: Adapter 1, Adapter 2, and Adapter 3. Adapter 1, which consists of two convolutional layers, channel weights, and attention, effectively learns new representations through fine-tuning. Adapters 2 and 3, which have fewer parameters than the decoder convolutional layer, reduce the number of fine-tuning parameters. Simulation results show that Adapter 1 improves the Normalized Mean Square Error (NMSE) between the estimated CSI and the reconstructed CSI by an average of 0.4 dB compared to the conventional multi-task learning-based CSI feedback. Adapters 2 and 3 reduce the number of parameters that need to be retained in the BS by 29% and 34%, respectively, compared to multi-task learning-based CSI feedback.
Keyword (in Japanese) (See Japanese page) 
(in English) CSI feedback / multi-task learning / fine-tuning / / / / /  
Reference Info. ITE Tech. Rep., vol. 48, no. 5, BCT2024-27, pp. 25-28, Feb. 2024.
Paper # BCT2024-27 
Date of Issue 2024-02-08 (BCT) 
ISSN Online edition: ISSN 2424-1970
Download PDF

Conference Information
Committee BCT IEEE-BT  
Conference Date 2024-02-15 - 2024-02-16 
Place (in Japanese) (See Japanese page) 
Place (in English) Nagoya International Center 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To BCT 
Conference Code 2024-02-BCT-BT 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Adapter-Based Fine-Tuning for Multi-Task Learning Based CSI Feedback in FDD Massive MIMO Systems 
Sub Title (in English)  
Keyword(1) CSI feedback  
Keyword(2) multi-task learning  
Keyword(3) fine-tuning  
1st Author's Name Mayuko Inoue  
1st Author's Affiliation Keio University (Keio Univ)
2nd Author's Name Tomoaki Ohtsuki  
2nd Author's Affiliation Keio University (Keio Univ)
3rd Author's Name  
3rd Author's Affiliation ()
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2024-02-16 10:50:00 
Presentation Time 20 minutes 
Registration for BCT 
Paper # BCT2024-27 
Volume (vol) vol.48 
Number (no) no.5 
Page pp.25-28 
Date of Issue 2024-02-08 (BCT) 

[Return to Top Page]

[Return to ITE Web Page]

The Institute of Image Information and Television Engineers (ITE), Japan