Institution profile

University of São Paulo

Academic institutionsouthamerica · br
Official website
Research library313linked papers
Opportunities0open roles
Selected work

Representative Papers

Cognitive Constructivism and the Epistemic Significance of Sharp Statistical Hypotheses in Natural Sciences

Jun 28, 2010

Traditional Bayesian paradigms neglect the ontological status of sharp statistical hypotheses, undermining their cognitive significance in natural science. Method: This study develops an epistemological framework aligned with cognitive constructivism, centering on the notion of “intrinsic solutions” to formalize the cognitive reality of sharp hypotheses. It introduces four criteria—sharpness, stability, separability, and composability—to assess their cognitive objectivity and grounds inference in the Full Bayesian Significance Test (FBST), moving beyond decision-theoretic Bayesian hypothesis testing. Contribution/Results: The work establishes the first operational, empirically verifiable constructivist statistical inference system, integrating FBST, subjectivist Bayesian modeling, formalized separability, and stability sensitivity analysis. It provides a novel, rigorous metric for evaluating the cognitive value of scientific hypotheses, thereby advancing both epistemology and statistical methodology.

23 citationsRead paper

Construction of an Impedance Control Test Bench

May 22, 2025Proceedings of the 27th International Congress of Mechanical Engineering

Robotic physical interaction suffers from coupled force–velocity control and a lack of unified experimental benchmarks. Method: This work introduces IC2D, an impedance control testbed featuring a novel electro-mechano-hydraulic compatible architecture—designed for high reliability and zero backlash—overcoming the limitation of conventional platforms that support only a single actuation modality. It enables hybrid electric/hydraulic actuation and combined linear/rotary motion. Through co-design across mechanical, electrical, and hydraulic domains; integration of multi-source high-precision sensing (force, position, velocity); real-time impedance/admittance control interfaces; and modular, reconfigurable kinematic structures, the platform enables quantitative assessment of joint-level interaction dynamics. Contribution/Results: IC2D establishes a state-of-the-art, multifunctional benchmarking system. It has successfully validated dynamic responses of diverse controllers and bandwidth characteristics of actuators, providing standardized experimental grounding for impedance control algorithm development and evaluation.

2 citationsRead paper

Combining YOLO and Visual Rhythm for Vehicle Counting

Nov 06, 2023Anais Estendidos da XXXVI Conference on Graphics, Patterns and Images (SIBRAPI Estendido 2023)

To address the high computational cost and poor real-time performance of vehicle counting in video surveillance, this paper proposes a single-stage, tracking-free counting method. The core innovation lies in the first integration of YOLO-based object detection with Visual Rhythm—a temporal-spatial imaging technique—that encodes vehicle trajectories into static images. Detection is performed only on adaptively selected keyframes, and counting is achieved directly via spatial pattern analysis of the resulting Visual Rhythm image. This paradigm eliminates the conventional two-stage “detection + tracking” pipeline, drastically reducing computational overhead. The method exhibits strong generalizability to arbitrary unidirectional motion counting scenarios. Evaluated on real-world video datasets, it achieves an average counting accuracy of 99.15% and processes videos three times faster than state-of-the-art tracking-based approaches.

2 citationsRead paper

Inference on model parameters with many L-moments

Oct 09, 2022

This paper addresses the low finite-sample efficiency of L-moment estimation arising from fixing the number of L-moments equal to the number of parameters. We propose an adaptive scheme wherein the number of L-moments—and their associated weights—increases with sample size. We establish, for the first time, an asymptotic theoretical framework for L-moment order selection that grows with sample size, and introduce the Generalized L-Moment Estimator (GLME). GLME retains asymptotic efficiency while substantially improving small-sample accuracy. Monte Carlo simulations demonstrate that GLME achieves significantly lower mean squared error than maximum likelihood estimation (MLE) in small samples, and converges asymptotically to MLE’s efficiency. An empirical application to Brazilian ride-hailing expenditure data further confirms its robustness and practical utility. The core innovation lies in relaxing the conventional fixed-dimension L-moment matching constraint, thereby unifying finite-sample superiority with asymptotic optimality.

2 citationsRead paper

Teaching Complex Systems based on Microservices

Jun 19, 2025

Undergraduate microservices education often lacks systematicity, conceptual depth, and hands-on rigor. Method: This study designed and implemented a structured microservices pedagogical framework for senior undergraduate computer science students at the University of São Paulo, involving over 80 learners. The framework integrates authentic industrial scenarios, collaborative learning, and agile development practices, covering core technical competencies—including microservices architecture, Docker-based containerization, RESTful API design, and CI/CD automation pipelines. Unlike conventional curricula, it delivers the first undergraduate-level, lifecycle-oriented instruction spanning microservices design, development, deployment, and team collaboration. Contribution/Results: Empirical evaluation demonstrated 100% project completion rate; students exhibited significantly enhanced comprehension of distributed systems principles and engineering proficiency; industry-readiness assessments improved by an average of 37%.

1 citationsRead paper
Recent publications

Latest Papers

It Is Not My Code Anymore

Sep 08, 2026

本文探讨了AI辅助编程中代码归属与责任问题，通过假设案例和文献分析，提出应基于服务结果设定接受标准，并讨论了质量工程在评估生成实现及过程中的作用。

0 citationsRead paper