| Creator |
|
|
|
|
|
| Language |
|
| Publisher |
|
| Date |
|
| Source Title |
|
| Vol |
|
| Issue |
|
| First Page |
|
| Last Page |
|
| Publication Type |
|
| Access Rights |
|
| JaLC DOI |
|
| Related DOI |
|
|
|
|
|
| Related URI |
|
|
|
|
|
| Relation |
|
|
|
|
|
| Abstract |
超並列パラダイムに基づく画像理解システムICEを提案し、その基本的な構成を示した。このシステムは階層的なニューラルネットワークをベースにしたものであり、本論文では、論理和的作用を行う特徴統合層と、論理積的作用を行う特徴抽出層のネットワークの構造を定めた。また、実験を通して、選択的注意機構、曖昧さの解消、欠損情報の補完などの画像認識に重要な機能が実現できることを示した。
In this paper, ...we propose the Image CEntered (ICE) System for image understanding. The system, based on a multi-layered iterative neural network, is performed in massively parallel fashion. In the ICE System, a result of image understanding is represented in a sequence of combinations of activated units ill the highest layer: each of the units corresponds to a word meaning. This is similar to the fact that a complex situation is represented in a sequence of simple sentences. In the lowest layer, a reconstructed image, common to all sub-networks which extract word meanings, is generated. The sub-networks can cooperate each other via the reconstructed image. This cooperative mechanism provides the system high performance and high flexibility of image understanding.show more
|
Export Mendeley
p237