Abstract:

Hashing techniques can effectively solve the storage and retrieval efficiency problems faced by Recommender Systems（RSs）. However, one issue of applying hashing to RSs is that RSs focus on modeling user’s preference over items rather than their similarities concerned by hashing. Therefore, an improved hashing for efficient recommendation method is proposed. The mean of each user and item relative scoring system is considered as a bias. The rating is mapped to the similarity interval by subtracting the bias term. For preserving the similarity mentioned above, two methods are proposed to decompose the similarity matrix to obtain user and item binary codes. Extensive experiments performed on three real-world benchmarks show that the method outperforms the state-of-the-art methods.

Key words: hashing, recommender systems, similarities, user’s preference, bias

摘要：

哈希技术能有效地解决推荐系统面临的存储和检索效率的问题。然而，现存的哈希推荐方法存在一个问题，推荐关注于建模用户对项目的偏好，而哈希学习关心的是相似性。为此，提出了一种改进的哈希推荐方法。计算每个用户、项目相对评分系统的均值作为偏置。对用户评分矩阵进行去偏置处理，将评分映射到相似性区间。以保持相似性为目标，提出了两种方式来分解相似性矩阵得到用户和项目的二进制码。在三个真实数据集上的实验结果表明，与其他方法对比，提出的方法在检索精度上有一定的优势。

关键词: 哈希学习, 推荐系统, 相似性, 用户偏好, 偏置