Institution profile

CINVESTAV

Academic institutionnorthamerica · mx
Official website
Research library24linked papers
Opportunities0open roles
Selected work

Representative Papers

SAGRAD: A Program for Neural Network Training with Simulated Annealing and the Conjugate Gradient Method

Jun 17, 2015Journal of Research of the National Institute of Standards and Technology

To address the non-convex optimization challenge in neural network classification—specifically, susceptibility to poor local minima and flat regions—this paper proposes SAGRAD, a batch-training algorithm integrating Simulated Annealing (SA) with Møller’s Scaled Conjugate Gradient (SCG) method. Its core innovation lies in the first incorporation of SA into the SCG framework, enabling a dynamic restart and escape mechanism that synergistically balances global exploration and local acceleration. Implemented in Fortran 77, SAGRAD incorporates efficient Hessian-vector multiplication, optimized gradient computation, and an adaptive SA weight initialization strategy. Empirical evaluation across multiple classification benchmarks demonstrates significantly improved convergence robustness and generalization performance, while markedly reducing the probability of converging to suboptimal local minima. These results validate SAGRAD’s effectiveness and practicality for non-convex optimization in neural network training.

6 citationsRead paper

OsteoCAD: A Human-in-the-Loop Cloud-Edge Framework for Bone Tumor Segmentation

Jul 31, 2026

This work addresses the challenge faced by resource-constrained medical institutions—limited computational capacity and insufficient expertise—that hinders the deployment of deep learning for bone tumor segmentation. To overcome this barrier, the authors propose a modular cloud-edge collaborative framework that encapsulates the entire pipeline, from data creation and preprocessing to model training and inference, into user-friendly interfaces through a human-in-the-loop architecture. By leveraging remote GPU resources, dynamic GPU scheduling, and an intuitive interactive interface within a cloud-edge computing paradigm, the system enables low-barrier AI deployment. Its efficacy is demonstrated in a real-world clinical setting in Mexico, where it successfully performed large-scale bone tumor segmentation, confirming the feasibility of deploying effective AI-driven medical solutions without requiring complex local infrastructure.

0 citationsRead paper

Human4K: A Large-Scale 4K Multi-View Mocap Dataset for Whole-Body 3D Human Reconstruction

Jul 15, 2026

Existing 3D human reconstruction methods often underperform in complex real-world scenarios due to geometric instability, inaccurate joint estimation, and self-occlusion, primarily caused by the lack of high-resolution, high-fidelity training data with diverse poses. To address this gap, this work introduces Human4K, a novel dataset that uniquely combines eight-view synchronized 4K video, professional Vicon motion capture, and highly self-occluded full-body motions across 11 subjects, yielding over six million frames. High-fidelity SMPL-X annotations are generated via a custom Motion Retargeting and Refinement Module (MRRM). Models trained on Human4K achieve significantly improved reconstruction accuracy on standard benchmarks, with notable gains in challenging regions such as hands, feet, and depth-ambiguous areas, thereby filling a critical void in high-fidelity full-body 3D human reconstruction data.

0 citationsRead paper
Recent publications

Latest Papers

OsteoCAD: A Human-in-the-Loop Cloud-Edge Framework for Bone Tumor Segmentation

Jul 31, 2026

This work addresses the challenge faced by resource-constrained medical institutions—limited computational capacity and insufficient expertise—that hinders the deployment of deep learning for bone tumor segmentation. To overcome this barrier, the authors propose a modular cloud-edge collaborative framework that encapsulates the entire pipeline, from data creation and preprocessing to model training and inference, into user-friendly interfaces through a human-in-the-loop architecture. By leveraging remote GPU resources, dynamic GPU scheduling, and an intuitive interactive interface within a cloud-edge computing paradigm, the system enables low-barrier AI deployment. Its efficacy is demonstrated in a real-world clinical setting in Mexico, where it successfully performed large-scale bone tumor segmentation, confirming the feasibility of deploying effective AI-driven medical solutions without requiring complex local infrastructure.

0 citationsRead paper

Human4K: A Large-Scale 4K Multi-View Mocap Dataset for Whole-Body 3D Human Reconstruction

Jul 15, 2026

Existing 3D human reconstruction methods often underperform in complex real-world scenarios due to geometric instability, inaccurate joint estimation, and self-occlusion, primarily caused by the lack of high-resolution, high-fidelity training data with diverse poses. To address this gap, this work introduces Human4K, a novel dataset that uniquely combines eight-view synchronized 4K video, professional Vicon motion capture, and highly self-occluded full-body motions across 11 subjects, yielding over six million frames. High-fidelity SMPL-X annotations are generated via a custom Motion Retargeting and Refinement Module (MRRM). Models trained on Human4K achieve significantly improved reconstruction accuracy on standard benchmarks, with notable gains in challenging regions such as hands, feet, and depth-ambiguous areas, thereby filling a critical void in high-fidelity full-body 3D human reconstruction data.

0 citationsRead paper

One country, multiple portraits: representativeness in GPS-based mobility data is source-specific and spatially dependent

Jun 22, 2026

This study addresses the pervasive issue of non-representative GPS mobility data in low- and middle-income countries, where coverage biases across data sources remain poorly understood. Integrating Facebook and Veraset GPS datasets with census data from 2,478 Mexican municipalities, the authors employ interpretable machine learning and spatial statistical models to systematically dissect the origins, spatial structure, and drivers of coverage bias. Findings reveal that coverage bias exhibits strong data-source specificity and spatial dependence: Facebook data demonstrate more uniform coverage, whereas multi-app aggregated data disproportionately represent wealthier, more digitally connected areas. Explicitly modeling spatial autocorrelation substantially improves explanatory power for these biases. The results underscore the necessity of tailoring bias-correction strategies to specific data sources and highlight that a portion of spatial variation in coverage cannot be fully accounted for by observable covariates.

0 citationsRead paper