Institution profile

Bucknell University

Academic institutionnorthamerica · us
Official website
Research library13linked papers
Opportunities0open roles
Selected work

Representative Papers

Improving TensorSketch Using Complex Random Variables

Aug 11, 2026

This work addresses the issue of exponentially growing variance in traditional TensorSketch when estimating high-order polynomial kernels, where the variance scales as $3^p/D$ with the degree $p$, severely degrading accuracy. The authors propose a novel TensorSketch variant based on complex-valued random variables, introducing complex random projections into the sparse framework of Pham et al. for the first time. This approach maintains the original time complexity of $O(p(\text{nnz}(x) + D \log D))$ while significantly reducing the variance bound to $2^p/D$. Both theoretical analysis and empirical evaluations demonstrate that the proposed method consistently improves estimation accuracy and computational efficiency across synthetic and real-world datasets.

0 citationsRead paper

Investigating Gender Bias in Touch Biometrics

Jun 09, 2026

This study investigates the presence of gender bias in swipe-based behavioral biometric authentication to ensure equitable performance across genders. Leveraging the BBMAS and ANTAL datasets, the authors employ XGBoost and DenseNet models and, for the first time in this domain, apply non-parametric statistical tests—including Kolmogorov-Smirnov, Mann-Whitney U, and Wasserstein permutation tests—to systematically evaluate gender disparities in authentication performance. Experimental results demonstrate that XGBoost achieves 92% and 94% accuracy on the two datasets, respectively, with no statistically significant differences in false acceptance rates (FAR) or false rejection rates (FRR) between male and female users across most configurations. These findings indicate that high authentication accuracy and low gender bias can be simultaneously attained.

0 citationsRead paper

Deterministic vs. LLM-Controlled Orchestration for COBOL-to-Python Modernization

May 10, 2026

This study addresses the challenges of COBOL system modernization—namely, the scarcity of domain experts, the scale of legacy codebases, and stringent correctness requirements—and investigates the efficacy of large language model (LLM)-based orchestration strategies in automated COBOL-to-Python translation. For the first time, orchestration strategy is isolated as the sole variable within a unified experimental framework, enabling a direct comparison between deterministic and LLM-driven approaches. The findings reveal that deterministic orchestration achieves functional correctness on par with LLM-based control while significantly enhancing robustness, reducing inter-run performance variability, and cutting token consumption by up to 3.5×. These results demonstrate that deterministic orchestration offers superior stability and cost efficiency without compromising translation accuracy.

0 citationsRead paper
Recent publications

Latest Papers

Improving TensorSketch Using Complex Random Variables

Aug 11, 2026

This work addresses the issue of exponentially growing variance in traditional TensorSketch when estimating high-order polynomial kernels, where the variance scales as $3^p/D$ with the degree $p$, severely degrading accuracy. The authors propose a novel TensorSketch variant based on complex-valued random variables, introducing complex random projections into the sparse framework of Pham et al. for the first time. This approach maintains the original time complexity of $O(p(\text{nnz}(x) + D \log D))$ while significantly reducing the variance bound to $2^p/D$. Both theoretical analysis and empirical evaluations demonstrate that the proposed method consistently improves estimation accuracy and computational efficiency across synthetic and real-world datasets.

0 citationsRead paper

Investigating Gender Bias in Touch Biometrics

Jun 09, 2026

This study investigates the presence of gender bias in swipe-based behavioral biometric authentication to ensure equitable performance across genders. Leveraging the BBMAS and ANTAL datasets, the authors employ XGBoost and DenseNet models and, for the first time in this domain, apply non-parametric statistical tests—including Kolmogorov-Smirnov, Mann-Whitney U, and Wasserstein permutation tests—to systematically evaluate gender disparities in authentication performance. Experimental results demonstrate that XGBoost achieves 92% and 94% accuracy on the two datasets, respectively, with no statistically significant differences in false acceptance rates (FAR) or false rejection rates (FRR) between male and female users across most configurations. These findings indicate that high authentication accuracy and low gender bias can be simultaneously attained.

0 citationsRead paper

Deterministic vs. LLM-Controlled Orchestration for COBOL-to-Python Modernization

May 10, 2026

This study addresses the challenges of COBOL system modernization—namely, the scarcity of domain experts, the scale of legacy codebases, and stringent correctness requirements—and investigates the efficacy of large language model (LLM)-based orchestration strategies in automated COBOL-to-Python translation. For the first time, orchestration strategy is isolated as the sole variable within a unified experimental framework, enabling a direct comparison between deterministic and LLM-driven approaches. The findings reveal that deterministic orchestration achieves functional correctness on par with LLM-based control while significantly enhancing robustness, reducing inter-run performance variability, and cutting token consumption by up to 3.5×. These results demonstrate that deterministic orchestration offers superior stability and cost efficiency without compromising translation accuracy.

0 citationsRead paper