arXiv subjects

Xiaoyao Yu

Publications and source records attributed to Xiaoyao Yu.

6 recordsLinked to original sources

arXiv2026

Position: AI for Science Should Treat Measurement-to-Dataset Pipelines as Inference Components

AI for Science (AI4Science) workflows often treat the released dataset as a fixed interface to the underlying system. However, in domains relying on \emph{indirect observation}, the learner observes a derivative representation produced by multi-stage measurement, reconstruction, and preprocessing pipelines. \textbf{We argue that these measurement-to-dataset pipelines are inference components: treating their outputs as ``given data'' freezes an observation model and obscures uncertainty over feasible pipeline choices.} We identify three failure modes arising from this ``frozen lens'': \textbf{(C1) hidden hypothesis space}, where the released dataset does not specify the pipeline configuration or its validity conditions; \textbf{(C2) uncertified transportability}, where a pipeline may be documented but its regime of validity is untested, so failures under distribution shift cannot be adjudicated; \textbf{(C3) ungoverned multiplicity}, where many defensible pipelines exist and dispersion is real but not propagated into uncertainty-aware evidence. We stress-test these claims with a large-scale neuroscience empirical audit, finding a survival rate of $\approx 0.0004\%$ under a cross-dataset stability criterion. We call on the AI4Science community to make pipelines \emph{computable} inference objects via domain-specific Computable Observation Frameworks. This shift enables quantifying pipeline adequacy and stability, converting implicit implementation choices into auditable, reproducible, and cumulative scientific evidence.

cs.LG↗

arXiv2025

MoE-Based Learned Inertial Odometry for Bicycle Localization

GNSS suffers from multipath errors in urban canyons, making reliable bicycle localization difficult. Hand-crafted inertial alternatives, such as cycling dead reckoning and nonholonomic constraints, fail to generalize across riders, postures, and road surfaces. This work implements learned inertial odometry for bicycle localization: a neural velocity predictor, trained on multirider, multi-bicycle, and multi-surface IMU data, is tightly coupled with an extended Kalman filter for three-dimensional pose estimation, replacing hand-crafted cycling models with a data-driven representation. To reduce computational overhead, the predictor is realized as a sparsely-gated Mixture of Experts (MoE) network with top-K routing, trained under an alternating parameter-freezing scheme with a per-expert capacity constraint to encourage diverse and balanced specialization.The proposed model attains 0.333 m/s inference error, 9.49 m ATE, and 2.58 m RTE at only 28.76 M FLOPs. This corresponds to approximately 7 times and 9 times fewer FLOPs than the ResMLP and ResNet backbones used in LLIO and TLIO, while maintaining comparable or superior accuracy. The proposed MoE variants further rank first on both ATE and RTE across all evaluation splits, demonstrating strong generalization.

cs.RO↗

arXiv2025

Beyond Pairwise Connections: Extracting High-Order Functional Brain Network Structures under Global Constraints

Functional brain network (FBN) modeling often relies on local pairwise interactions, whose limitation in capturing high-order dependencies is theoretically analyzed in this paper. Meanwhile, the computational burden and heuristic nature of current hypergraph modeling approaches hinder end-to-end learning of FBN structures directly from data distributions. To address this, we propose to extract high-order FBN structures under global constraints, and implement this as a Global Constraints oriented Multi-resolution (GCM) FBN structure learning framework. It incorporates 4 types of global constraint (signal synchronization, subject identity, expected edge numbers, and data labels) to enable learning FBN structures for 4 distinct levels (sample/subject/group/project) of modeling resolution. Experimental results demonstrate that GCM achieves up to a 30.6% improvement in relative accuracy and a 96.3% reduction in computational time across 5 datasets and 2 task settings, compared to 9 baselines and 10 state-of-the-art methods. Extensive experiments validate the contributions of individual components and highlight the interpretability of GCM. This work offers a novel perspective on FBN structure learning and provides a foundation for interdisciplinary applications in cognitive neuroscience. Code is publicly available on https://github.com/lzhan94swu/GCM.

cs.LG↗

arXiv2025

Knowledge Independence Breeds Disruption but Limits Recognition

Despite extensive research on scientific disruption, two questions remain: why disruption has declined amid growing knowledge, and why disruptive work receives fewer and delayed citations. One way to address these questions is to identify an intrinsic, paper-level property that reliably predicts disruption and explains both patterns. Here, we propose a novel measure, knowledge independence, capturing the extent to which a paper draws on references that do not cite one another. Analyzing 114 million publications, we find that knowledge independence strongly predicts disruption and mediates the disruptive advantage of small, onsite, and fresh teams. Its long-term decline, nonreproducible by null models, provides a mechanistic explanation for the parallel decline in disruption. Causal and simulation evidence further indicates that knowledge independence drives the persistent trade-off between disruption and impact. Taken together, these findings fill a critical gap in understanding scientific innovation, revealing a universal law: Knowledge independence breeds disruption but limits recognition.

physics.soc-ph↗

arXiv2021

Become a better you: correlation between the change of research direction and the change of scientific performance

It is important to explore how scientists decide their research agenda and the corresponding consequences, as their decisions collectively shape contemporary science. There are studies focusing on the overall performance of individuals with different problem choosing strategies. Here we ask a slightly different but relatively unexplored question: how is a scientist's change of research agenda associated with her change of scientific performance. Using publication records of over 14,000 authors in physics, we quantitatively measure the extent of research direction change and the performance change of individuals. We identify a strong positive correlation between the direction change and impact change. Scientists with a larger direction change not only are more likely to produce works with increased scientific impact compared to their past ones, but also have a higher growth rate of scientific impact. On the other hand, the direction change is not associated with productivity change. Those who stay in familiar topics do not publish faster than those who venture out and establish themselves in a new field. The gauge of research direction in this work is uncorrelated with the diversity of research agenda and the switching probability among topics, capturing the evolution of individual careers from a new point of view. Though the finding is inevitably affected by the survival bias, it sheds light on a range of problems in the career development of individual scientists.

physics.soc-ph↗

arXiv2021

The network asymmetry caused by the degree correlation and its effect on the bimodality in control

Our ability to control a whole network can be achieved via a small set of driver nodes. While the minimum number of driver nodes needed for control is fixed in a given network, there are multiple choices for the driver node set. A quantity used to investigate this multiplicity is the fraction of redundant nodes in the network, referring to nodes that do not need any external control. Previous work has discovered a bimodality feature characterized by a bifurcation diagram: networks with the same statistical property would stay with equal probability to have a large or small fraction of redundant nodes. Here we find that this feature is rooted in the symmetry of the directed network, where both the degree distribution and the degree correlation can play a role. The in-in and out-out degree correlation will suppress the bifurcation, as networks with such degree correlations are asymmetric under network transpose. The out-in and in-out degree correlation do not change the network symmetry, hence the bimodality feature is preserved. However, the out-in degree correlation will change the critical average degree needed for the bifurcation. Hence by fixing the average degree of networks and tuning out-in degree correlation alone, we can observe a similar bifurcation diagram. We conduct analytical analyses that adequately explain the emergence of bimodality caused by out-in degree correlation. We also propose a quantity, taking both degree distribution and degree correlation into consideration, to predict if a network would be at the upper or lower branch of the bifurcation. As is well known that most real networks are not neutral, our results extend our understandings of the controllability of complex networks.

physics.soc-ph↗