25. 基本原理
movie
person W X Y Z
A 5 4
B 4 3 3
C 3
D 4
E 1 5
列ベクトルについて、一番近いものを探す
「映画Xが好きな人は、映画Wも好き」
26. 全体構成
入力:大規模な疎行列
対CPU比計算速度
特異値分解
(SVD) x30
k最近傍探索
(kNN) x130
我々の成果!
(注:CPUシングルコアとの比較)
論文
K.Kato and T.Hosino, Solving k-Nearest Neighbor Problem on Multiple Graphics Processors, In Proc. CCGrid2010,
Melbourne, Australia, pp 769-773, 2010.
K.Kato and T.Hosino, Singular Value Decomposition for Collaborative Filtering on a GPU, IOP Conference Series:
Materials Science and Engineering 10 012017, 2010.
K.Kato and T.Hosino. Multi-GPU algorithm for k-nearest neighbor problem. Concurrency and Computation: Practice
and Experience, 23, 2011.
31. 解決策
次元方向に刻んで、スライスごとに共有メモリに読み込む
Block 0
Block 1
Block 2
K.Kato and T.Hosino, Solving k-Nearest Neighbor Problem on Multiple Graphics Processors,
In Proc. CCGrid2010, Melbourne, Australia, pp 769-773, 2010.
K.Kato and T.Hosino. Multi-GPU algorithm for k-nearest neighbor problem. Concurrency and
Computation: Practice and Experience, 23, 2011.