Development of advanced inference technologies for huge knowledge graphs in tensor spaces
Project/Area Number |
18H03288
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Review Section |
Basic Section 61030:Intelligent informatics-related
|
Research Institution | Wakayama University |
Principal Investigator |
Sakama Chiaki 和歌山大学, システム工学部, 教授 (20273873)
|
Co-Investigator(Kenkyū-buntansha) |
井上 克巳 国立情報学研究所, 情報学プリンシプル研究系, 教授 (10252321)
林 克彦 東京大学, 大学院情報理工学系研究科, 助教 (50725794)
|
Project Period (FY) |
2018-04-01 – 2022-03-31
|
Project Status |
Completed (Fiscal Year 2022)
|
Budget Amount *help |
¥17,030,000 (Direct Cost: ¥13,100,000、Indirect Cost: ¥3,930,000)
Fiscal Year 2021: ¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2020: ¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2019: ¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2018: ¥4,940,000 (Direct Cost: ¥3,800,000、Indirect Cost: ¥1,140,000)
|
Keywords | テンソル空間 / 線形代数計算 / 論理プログラミング / 演繹推論 / デフォルト推論 / アブダクション / 論理推論 / スパース行列 / 部分計算 / 知識グラフ / 高次推論 / 近似計算 / 人工知能 / 量子化 |
Outline of Final Research Achievements |
In this study, we proposed a new efficient computation method for advanced reasoning in artificial intelligence on a large-scale knowledge base such as a knowledge graph, which is different from conventional theorem proving methods. More precisely, we developed a theory for converting a knowledge base represented as a logic program into an algebraic representation in tensor space, and computing deductive inference, default inference, and abduction using linear algebra. Next we implemented the system and conducted experimental evaluations using both artificial and real data. Furthermore, we developed optimization techniques for computational speed-up and confirmed their effectiveness in experimental verification. The proposed method has the potential to dramatically improve the scalability of symbolic reasoning through parallel algebraic computation using GPUs.
|
Academic Significance and Societal Importance of the Research Achievements |
ニューラルネットワーク(NN)による機械学習の結果に説明可能性を与え信頼性を向上させるための手法として、NNと記号処理を融合したニューロシンボリックAIが提案されている。しかし、従来の記号処理計算を大規模データに適用する場合、計算効率がボトルネックとなる問題があった。本研究で提案したテンソル空間における高次推論技術は、NNで扱われているような高次元のベクトル表現されたデータを使った推論を線形代数的に計算することを可能にする。その学術的意義は、大規模データからのスケーラブルで高速な推論を実現するための新しい計算技術の提案であり、社会的意義はNNと記号処理の融合に向けた要素技術としての貢献である。
|
Report
(5 results)
Research Products
(30 results)
-
-
-
-
-
-
-
-
-
-
[Journal Article] Binarized Knowledge Graph Embeddings2019
Author(s)
Koki Kishimoto, Katsuhiko Hayashi, Genki Akai, Masashi Shimbo, Kazunori Komatani
-
Journal Title
Advances in Information Retrieval. ECIR 2019. Lecture Notes in Computer Science
Volume: 11437
Pages: 181-196
DOI
ISBN
9783030157111, 9783030157128
Related Report
Peer Reviewed
-
[Journal Article] Computing Logic Programming Semantics in Linear Algebra2018
Author(s)
Hien D. Nguyen, Chiaki Sakama, Taisuke Sato, Katsumi Inoue
-
Journal Title
In: Proceedings of the 12th International Conference on Multi-disciplinary Trends in Artificial Intelligence (MIWAI 2018), Lecture Notes in Computer Science
Volume: 11248
Pages: 32-48
DOI
ISBN
9783030030131, 9783030030148
Related Report
Peer Reviewed / Int'l Joint Research
-
-
[Journal Article] Partial Evaluation of Logic Programs in Vector Spaces2018
Author(s)
Chiaki Sakama, Hien D. Nguyen, Taisuke Sato, Katsumi Inoue
-
Journal Title
Proceedings of the 11th International Workshop on Answer Set Programming and Other Computing Paradigms (ASPOCP 2018), CoRR abs
Volume: 1811.11435
Pages: 1-14
Related Report
Peer Reviewed / Open Access / Int'l Joint Research
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-