Selected work

Representative Papers

Solution for 10th Competition on Ambivalence/Hesitancy (AH) Video Recognition Challenge using Divergence-Based Multimodal Fusion

Mar 15, 2026

This study addresses the challenge of recognizing ambivalence/hesitancy (A/H) states in video-recorded individuals by proposing a multimodal fusion approach based on inter-modal embedding discrepancies. The method explicitly models cross-modal inconsistency—a hallmark of A/H—by computing absolute differences among visual, audio, and textual embeddings in a shared latent space. Visual features are derived using Py-Feat to extract action units (AUs), audio representations are obtained via Wav2Vec 2.0, and textual inputs are encoded with BERT; these modalities are subsequently aligned through a BiLSTM combined with attention-based pooling. Evaluated on the BAH dataset, the model achieves a Macro F1 score of 0.6808, substantially outperforming baseline methods (0.2827). Further analysis confirms that temporal variability in AUs serves as a critical visual discriminative cue for A/H detection.

0 citationsRead paper

BROTHER: Behavioral Recognition Optimized Through Heterogeneous Ensemble Regularization for Ambivalence and Hesitancy

Mar 15, 2026

This work addresses the challenge of recognizing complex behaviors such as contradiction and hesitation in naturalistic videos, which manifest through subtle and multimodal conflicts. The authors propose a regularized multimodal fusion framework that incorporates a statistical textual modality to capture temporal dynamics of speech, alongside visual and acoustic features. A heterogeneous model selection mechanism calibrated by validation loss is introduced, followed by a hard-voting ensemble strategy optimized via particle swarm optimization (PSO) with a penalty term (λ=0.2) on the training-validation gap to mitigate overfitting. Experimental results demonstrate that the linguistic modality serves as the strongest individual predictor, and the proposed approach achieves a Macro F1-score of 0.7465 on an unseen test set, significantly enhancing generalization performance.

0 citationsRead paper

Assessing Redundancy Strategies to Improve Availability in Virtualized System Architectures

Nov 25, 2025

To address insufficient availability of Nextcloud file servers in private cloud environments, this paper proposes a dual-redundancy architecture jointly operating at the host and virtual machine layers. A system reliability model is constructed using Stochastic Petri Nets (SPNs) and quantitatively evaluated on an Apache CloudStack-based private cloud platform. Compared to single-layer redundancy strategies, the proposed approach significantly improves steady-state system availability, reducing expected downtime by up to 42.6%. The key contributions are: (i) the first formal modeling of cross-layer dual redundancy as a coupled SPN, enabling integrated characterization of inter-layer fault propagation and recovery dynamics; and (ii) a verifiable, reusable modeling framework and decision-support methodology for designing highly available virtualized private cloud infrastructures.

0 citationsRead paper

Multitask LSTM for Arboviral Outbreak Prediction Using Public Health Data

May 07, 2025

This work addresses the challenge of jointly forecasting dengue, chikungunya, and Zika—three arboviral diseases—in Recife, Brazil. We propose the first unified multi-task temporal modeling framework specifically designed for arboviruses. Leveraging public health data from DataSUS (2017–2023), we develop an LSTM-based multi-task model that simultaneously performs outbreak detection (binary classification) and case count prediction (regression). The framework incorporates sliding temporal windows, hyperparameter optimization via Keras Tuner, and time-series cross-validation. Key findings indicate that a 90-day window achieves optimal trade-off between classification and regression performance. On an independent 2023 test set, the model reduces dengue case prediction MAE by 18.7% and achieves an F1-score of 0.82 for outbreak classification. The framework significantly enhances generalizability and task synergy in data-scarce settings.

0 citationsRead paper

An analysis of data variation and bias in image-based dermatological datasets for machine learning classification

Jan 15, 2025

Clinical smartphone images exhibit significant distribution shifts relative to professional dermoscopic images—across illumination, skin tone, viewpoint, noise, and annotation quality—degrading the generalizability of AI-based skin cancer classifiers. Method: We quantitatively characterize this domain shift for the first time and propose a small-sample cross-distribution data fusion strategy. Through distribution shift analysis, domain adaptation evaluation, multi-architecture benchmarking, class imbalance correction, and transfer learning diagnostics, we identify skin tone bias and low-resolution aliasing as primary confounding factors. Contribution/Results: Our hybrid training approach improves classification accuracy by up to 12.3% on real-world clinical smartphone images, substantially mitigating performance degradation during domain transfer. The methodology provides a reproducible, robust framework for deploying medical AI in resource-constrained clinical settings.

0 citationsRead paper

Recent publications

Latest Papers

Solution for 10th Competition on Ambivalence/Hesitancy (AH) Video Recognition Challenge using Divergence-Based Multimodal Fusion

Mar 15, 2026

This study addresses the challenge of recognizing ambivalence/hesitancy (A/H) states in video-recorded individuals by proposing a multimodal fusion approach based on inter-modal embedding discrepancies. The method explicitly models cross-modal inconsistency—a hallmark of A/H—by computing absolute differences among visual, audio, and textual embeddings in a shared latent space. Visual features are derived using Py-Feat to extract action units (AUs), audio representations are obtained via Wav2Vec 2.0, and textual inputs are encoded with BERT; these modalities are subsequently aligned through a BiLSTM combined with attention-based pooling. Evaluated on the BAH dataset, the model achieves a Macro F1 score of 0.6808, substantially outperforming baseline methods (0.2827). Further analysis confirms that temporal variability in AUs serves as a critical visual discriminative cue for A/H detection.

0 citationsRead paper

BROTHER: Behavioral Recognition Optimized Through Heterogeneous Ensemble Regularization for Ambivalence and Hesitancy

Mar 15, 2026

This work addresses the challenge of recognizing complex behaviors such as contradiction and hesitation in naturalistic videos, which manifest through subtle and multimodal conflicts. The authors propose a regularized multimodal fusion framework that incorporates a statistical textual modality to capture temporal dynamics of speech, alongside visual and acoustic features. A heterogeneous model selection mechanism calibrated by validation loss is introduced, followed by a hard-voting ensemble strategy optimized via particle swarm optimization (PSO) with a penalty term (λ=0.2) on the training-validation gap to mitigate overfitting. Experimental results demonstrate that the linguistic modality serves as the strongest individual predictor, and the proposed approach achieves a Macro F1-score of 0.7465 on an unseen test set, significantly enhancing generalization performance.

0 citationsRead paper

Assessing Redundancy Strategies to Improve Availability in Virtualized System Architectures

Nov 25, 2025

To address insufficient availability of Nextcloud file servers in private cloud environments, this paper proposes a dual-redundancy architecture jointly operating at the host and virtual machine layers. A system reliability model is constructed using Stochastic Petri Nets (SPNs) and quantitatively evaluated on an Apache CloudStack-based private cloud platform. Compared to single-layer redundancy strategies, the proposed approach significantly improves steady-state system availability, reducing expected downtime by up to 42.6%. The key contributions are: (i) the first formal modeling of cross-layer dual redundancy as a coupled SPN, enabling integrated characterization of inter-layer fault propagation and recovery dynamics; and (ii) a verifiable, reusable modeling framework and decision-support methodology for designing highly available virtualized private cloud infrastructures.

0 citationsRead paper

Multitask LSTM for Arboviral Outbreak Prediction Using Public Health Data

May 07, 2025

This work addresses the challenge of jointly forecasting dengue, chikungunya, and Zika—three arboviral diseases—in Recife, Brazil. We propose the first unified multi-task temporal modeling framework specifically designed for arboviruses. Leveraging public health data from DataSUS (2017–2023), we develop an LSTM-based multi-task model that simultaneously performs outbreak detection (binary classification) and case count prediction (regression). The framework incorporates sliding temporal windows, hyperparameter optimization via Keras Tuner, and time-series cross-validation. Key findings indicate that a 90-day window achieves optimal trade-off between classification and regression performance. On an independent 2023 test set, the model reduces dengue case prediction MAE by 18.7% and achieves an F1-score of 0.82 for outbreak classification. The framework significantly enhances generalizability and task synergy in data-scarce settings.

0 citationsRead paper

An analysis of data variation and bias in image-based dermatological datasets for machine learning classification

Jan 15, 2025

Clinical smartphone images exhibit significant distribution shifts relative to professional dermoscopic images—across illumination, skin tone, viewpoint, noise, and annotation quality—degrading the generalizability of AI-based skin cancer classifiers. Method: We quantitatively characterize this domain shift for the first time and propose a small-sample cross-distribution data fusion strategy. Through distribution shift analysis, domain adaptation evaluation, multi-architecture benchmarking, class imbalance correction, and transfer learning diagnostics, we identify skin tone bias and low-resolution aliasing as primary confounding factors. Contribution/Results: Our hybrid training approach improves classification accuracy by up to 12.3% on real-world clinical smartphone images, substantially mitigating performance degradation during domain transfer. The methodology provides a reproducible, robust framework for deploying medical AI in resource-constrained clinical settings.

0 citationsRead paper