The rapid accumulation of multi-modal data (e.g., text, images, and geo-locations) presents significant opportunities for data mining applications in healthcare and e-commerce. However, effectively retrieving such data remains challenging due to the difficulty in capturing diverse and dynamic user retrieval intents. Existing solutions, such as vector databases, typically rely on static embeddings or fixed weights, failing to adapt to users' varying preferences across different modalities. To address this, we present OneDB, a distributed framework for multi-metric similarity search that integrates data management with learning-based techniques. Unlike traditional systems, OneDB features three key algorithmic innovations: (i) an adaptive metric weight learning model based on lightweight contrastive learning, which infers implicit user preferences from limited query examples; (ii) a dual-layer indexing strategy that combines global partitioning with modality-aware local indexing to handle heterogeneous data distributions efficiently; and (iii) an end-to-end parameter tuning module leveraging deep reinforcement learning to optimize system performance in dynamic environments. Extensive experiments on real-world datasets demonstrate that OneDB captures user intent effectively, achieving 12.63%--30.75% higher accuracy and 2.5--5.75× faster retrieval speeds compared to state-of-the-art vector search systems.
Tang Qian, Yifan Zhu, Lu Chen et al.· Proceedings of the 32nd ACM...· 0 citations
Computational prediction of crystal properties plays a pivotal role in materials science. With the accelerated progress in machine learning, crystal property prediction has seen remarkable advancements. Nevertheless, the utilization of machine learning in this context faces several challenges. First, existing methods that utilize the smallest repeatable unit cell of a crystal often have a limited receptive field. Second, as experiments measuring crystal properties are time-consuming, labeled data is often scarce. To address these challenges, we propose a
S
elect
I
ve mu
L
ti-
V
iew representation
A
ugmentation framework (SILVA) for crystal property prediction. To go beyond limited receptive fields, we introduce the notion of a crystal supercell, which enables more comprehensive explorations of crystal structure. To fully combine insights from multi-view structures, i.e., from unit cells and supercells, we propose a multi-view representation learning (MRL) module that features a representation space that enhances the learning of representative features specific to different views. To alleviate the limited availability of labeled data, we propose a selective representation augmentation (SRA) module. Given representations of labeled training data, we carefully select nearby representations in the representation space established by the MRL module so that labels can be reused. An experimental study offers evidence that SILVA is capable of outperforming state-of-the-art methods.
Haomin Yu, Jilin Hu, K. Tolborg et al.· ACM Transactions on AI for S...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
