Imperial College London（帝国理工学院）

2503.21473 2026-04-01 stat.ML cs.LG

DeepRV: Accelerating Spatiotemporal Inference with Pre-trained Neural Priors

DeepRV: 通过预训练神经先验加速时空推断

Jhonathan Navott, Daniel Jenson, Seth Flaxman, Elizaveta Semenova

机构 * School of Public Health, Imperial College London（帝国理工学院公共卫生学院）； Department of Computer Science, University of Oxford（牛津大学计算机科学系）

AI总结 DeepRV通过预训练神经网络替代传统高斯过程先验，实现O(N²)复杂度，提升大规模时空数据推断效率与精度，适用于概率编程管道。

Comments Code to reproduce all experiments is available in the dl4bi codebase: https://github.com/MLGlobalHealth/dl4bi

2510.14710 2026-04-01 math.AT cs.LG physics.data-an

MCbiF: Measuring Topological Autocorrelation in Multiscale Clusterings via 2-Parameter Persistent Homology

MCbiF：通过2参数持续同调测量多尺度聚类的拓扑自相关

Juni Schindler, Mauricio Barahona

机构 * Department of Mathematics, Imperial College London, UK（英国伦敦帝国理工学院数学系）； Department of Mathematical Modeling and Machine Learning, University of Zurich, Switzerland（瑞士苏黎世大学数学建模与机器学习系）

AI总结本文提出MCbiF方法，利用2参数持续同调分析多尺度聚类序列的拓扑自相关，展示了其在非层次聚类任务中的有效性。

Comments Published as a conference paper at 14th International Conference on Learning Representations (ICLR 2026): https://openreview.net/forum?id=E7D6uybODJ

2510.12405 2026-04-01 cs.LG cond-mat.mtrl-sci

Continuous SUN (Stable, Unique, and Novel) Metric for Generative Modeling of Inorganic Crystals

连续SUN（稳定、唯一、新颖）度量用于无机晶体生成建模

Masahiro Negishi, Hyunsoo Park, Kinga O. Mastej, Aron Walsh

机构 * Imperial College London（伦敦帝国学院）

AI总结本文提出连续SUN度量，解决传统二元指标在晶体生成中的局限性，通过连续化提升评估精度与可调性，实验表明其能有效识别优质样本并避免强化学习中的奖励黑客问题。

Comments 23 pages (17 pages of main text). See https://github.com/WMD-group/xtalmet for the code. Significantly extended from the early version of this work, which was accepted to the AI4Mat workshop at NeurIPS 2025