2022 Fiscal Year Final Research Report
Efficient Lossy Compression for Non-Uniform Information Sources
| Project/Area Number |
17K00023
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Theory of informatics
|
|---|
| Research Institution | Kwansei Gakuin University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2017-04-01 – 2023-03-31
|
| Keywords | 非可逆データ圧縮 / 符号化 / 非一様情報源 / トレリス符号 / 確率伝搬法 |
|---|
| Outline of Final Research Achievements |
We studied lossy source compression in which original data is recovered from compressed data under certain fidelity criterion. In the first part of this project, we had emphasis on the design of trellis codes for non-uniformly distributed sources and non-binary convolutional codes with large free distance. Subsequently, we discussed the enhancement of algorithms for lossy compression in terms of the fidelity or computational and communication cost through the use of machine learning techniques, including the training of message-passing encoder for sparse graph source codes, and distributed recovery of sparse vectors observed by multiple sensing nodes.
|
| Free Research Field |
情報理論とその応用
|
| Academic Significance and Societal Importance of the Research Achievements |
ネットワーク上で送受信されるデータ量の増大に伴い，情報伝送の効率化が重要な課題となっており，その解決に向けて非可逆データ圧縮の高度化が有用となる．原情報に対する復元データの忠実度と，圧縮率の間には原理的にトレードオフが存在し，その理論的性能限界に迫る圧縮法の具体化が求められる．本研究では，非可逆データ圧縮に適したトレリス符号の構成，多元アルファベット上の畳込み符号の設計，機械学習に基づく符号化器の効率化，複数の観測ノードによる原情報の分散再構成に関する成果を得ており，理論・実用の双方の観点で情報伝送の効率化に寄与している．
|
