Strength Prediction and Experimental Allocation for Stabilized Moraine Using Gaussian Processes

Binder selection for surface moraine consolidation requires strength measurements across dosages and curing ages before injection trials. Twenty binder–dosage formulations were tested at four ages. Independent-task and multi-task Gaussian processes (IGP and MTGP) were evaluated for predicting untested formulations and allocating a limited testing budget. When only one formulation was withheld, IGP and linear interpolation had nearly identical mean absolute errors. Sharing information across binders improved prediction in some sparse designs but increased error when an entire dosage was withheld. MTGP produced narrower intervals with poorer coverage. Calibration improved simultaneous coverage across all four ages at the cost of substantially greater width. In retrospective sequential replay, integrated variance reduction (IVR) lowered the budget-averaged IGP prediction error by about 3.6% relative to random selection. Space-filling and maximum-variance selection achieved similar errors. Excluding the queried formulation from the IVR utility removed its advantage over random selection. On this regular dosage grid, measuring an underrepresented dosage was as effective as selecting it by covariance. The benefits of cross-binder sharing for sparse point prediction did not consistently carry over to interval quality or test allocation.

Authors

Institutions

Publication Details

Journal
Applied Sciences
Published
2026-11-18
DOI
https://doi.org/10.3390/app162211427
Primary Topic
Gaussian Processes and Bayesian Inference
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Strength Prediction and Experimental Allocation for Stabilized Moraine Using Gaussian Processes

Yalei Zhe, Kepeng Hou, Qunzhi Cheng, Guangquan Li et al.
1 links
Applied Sciences
Gaussian Processes and Bayesian Inference
article

Strength Prediction and Experimental Allocation for Stabilized Moraine Using Gaussian Processes

Yalei Zhe, Kepeng Hou, Qunzhi Cheng, Guangquan Li, Mingwu Liu, Chao Hong
article en

Abstract

Binder selection for surface moraine consolidation requires strength measurements across dosages and curing ages before injection trials. Twenty binder–dosage formulations were tested at four ages. Independent-task and multi-task Gaussian processes (IGP and MTGP) were evaluated for predicting untested formulations and allocating a limited testing budget. When only one formulation was withheld, IGP and linear interpolation had nearly identical mean absolute errors. Sharing information across binders improved prediction in some sparse designs but increased error when an entire dosage was withheld. MTGP produced narrower intervals with poorer coverage. Calibration improved simultaneous coverage across all four ages at the cost of substantially greater width. In retrospective sequential replay, integrated variance reduction (IVR) lowered the budget-averaged IGP prediction error by about 3.6% relative to random selection. Space-filling and maximum-variance selection achieved similar errors. Excluding the queried formulation from the IVR utility removed its advantage over random selection. On this regular dosage grid, measuring an underrepresented dosage was as effective as selecting it by covariance. The benefits of cross-binder sharing for sparse point prediction did not consistently carry over to interval quality or test allocation.

Applied SciencesVol. 16(22)
Kunming University of Science and Technology (CN)
Yunnan Provincial Science and Technology Department
Gaussian Processes and Bayesian Inference
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon