HGHuanying Gu

Huanying Gu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Huanying Gu has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    LossTransform: Reformulating the Loss Function for Contrastive Learning

    by Zheng Li, Jerry Cheng, Huanying Helen Gu

    Contrastive learning improves model performance by differentiating between positive and negative sample pairs. However, its application is primarily confined to classification tasks, facing challenges with complex recognition tasks such as object detection and segmentation due to its limited capacity to capture spatial relationships and fine-grained features. To address this limitation, we propose LossTransform, a novel approach that redefines positive sample pairs and establishes a novel contrastive loss paradigm. LossTransform advances contrastive learning to the instance level, departing from the traditional sample level. Empirical evaluations on ImageNet, CIFAR, and object detection benchmarks indicate that LossTransform improves accuracy by +2.73% on CIFAR, +2.52% on ImageNet, and up to +5.2% in average precision on detection tasks, while maintaining efficiency. These results illustrate that LossTransform is compatible with large-scale training pipelines and exhibits robust performance across diverse and complex datasets. By optimizing model performance and significantly reducing training time, this research enables more efficient and accessible solutions for societal applications.

  • Open Access

    Multimodal Models in Healthcare: Methods, Challenges, and Future Directions for Enhanced Clinical Decision Support

    by Kamrul Siam, Jobair Hossain Faruk, Bofan He, Jerry Q. Cheng, Huanying Gu

    Decision-making in modern healthcare increasingly relies on integrating a variety of data sources, including patient demographics, medical imaging, laboratory results, clinical narratives, and temporal data, all of which are difficult for traditional computational methodologies to accurately predict. This paper evaluates the latest methodologies that integrate diverse data types, including photographs, clinical notes, temporal measurements, and structured tables, through techniques such as feature amalgamation, prioritization of essential information, and utilization of graphs. We also assess pre-training, fine-tuning, and comprehensive evaluation of model generation procedures. By synthesizing findings from 50 of 91 peer-reviewed papers published between 2020 and 2024, we demonstrate that the integration of structured and unstructured data significantly improves performance in tasks like diagnosis, prognosis prediction, and personalized treatment. This review combines substantial multimodal datasets and applications across several therapeutic domains while addressing critical issues such as data heterogeneity, scalability, interpretability, and ethical considerations. This paper highlights the transformative potential of multimodal models in improving clinical decision support, providing a framework for future research to advance precision medicine and enhance healthcare outcomes.


  • Open Access

    A Globally Optimal Alternative to MLP

    by Zheng Li, Jerry Cheng, Huanying Helen Gu

    In deep learning, achieving the global minimum poses a significant challenge, even for relatively simple architectures such as Multi-Layer Perceptrons (MLPs). To address this challenge, we visualized model states at both local and global optima, thereby identifying the factors that impede the transition of models from local to global minima when employing conventional model training methodologies. Based on these insights, we propose the Lagrange Regressor (LReg), a framework that is mathematically equivalent to MLPs. Rather than updates via optimization techniques, LReg employs a Mesh-Refinement–Coarsening (discrete) process to ensure the convergence of the model’s loss function to the global minimum. LReg achieves faster convergence and overcomes the inherent limitations of neural networks in fitting multi-frequency functions. Experiments conducted on large-scale benchmarks including ImageNet-1K (image classification), GLUE (natural language understanding), and WikiText (language modeling) show that LReg consistently enhances the performance of pre-trained models, significantly lowers test loss, and scales effectively to big data scenarios. These results underscore LReg’s potential as a scalable, optimization-free alternative for deep learning in large and complex datasets, aligning closely with the goals of innovative big data analytics.

  • Open Access

    Sequential Policy Gradient for Adaptive Hyperparameter Optimization

    by Zheng Li, Jerry Cheng, Huanying Helen Gu

    Reinforcement learning is essential for neural architecture search and hyperparameter optimization, but the conventional approaches impede widespread use due to prohibitive time and computational costs. Inspired by DeepSeek-V3 multi-token prediction architecture, we propose Sequential Policy Gradient modeling (SPG), a novel trajectory generation paradigm for lightweight online hyperparameter optimization. In contrast to conventional policy gradient methods, SPG extends the base model with temporary modules, enabling it to generate state-action (padded) trajectories in a single forward pass. Our experiments demonstrate that models gain performance when retrained with SPG on their original datasets and also outperform standard transfer fine-tuning. We evaluate on five datasets spanning computer vision (ImageNet, COCO), natural language processing (GLUE, SQuAD), and audio (SUPERB) to assess the industrial applicability of SPG. The proposed method demonstrates consistent improvements across widely adopted models, achieving performance gains of $+0.2\sim7\%$, with significantly low computational costs. Fully reproducible code and pre-trained models: https://huggingface.co/UniversalAlgorithmic/SPG.

  • Open Access

    A Globally Optimal Alternative to MLP

    by Zheng Li, Jerry Cheng, Huanying Gu

    In deep learning, achieving the global minimum poses a significant challenge, even for relatively simple architectures such as Multi-Layer Perceptrons (MLPs). To address this challenge, we visualized model states at both local and global optima, thereby identifying the factors that impede the transition of models from local to global minima when employing conventional model training methodologies. Based on these insights, we propose the Lagrange Regressor (LReg), a framework that is mathematically equivalent to MLPs. Rather than updates via optimization techniques, LReg employs a mesh-refinement-coarsening (discrete) process to ensure the convergence of the model’s loss function to the global minimum. LReg achieves faster convergence and overcomes the inherent limitations of neural networks in fitting multi-frequency functions. Experiments conducted on large-scale benchmarks including ImageNet-1K (image classification), GLUE (natural language understanding), and WikiText (language modeling) show that LReg consistently enhances the performance of pre-trained models, significantly lowers test loss, and scales effectively to big data scenarios. These results underscore LReg’s potential as a scalable, optimization-free alternative for deep learning in large and complex datasets, aligning closely with the goals of innovative big data analytics.

  • Open Access

    LossTransform: Reformulating the Loss Function for Contrastive Learning

    by Zheng Li, Jerry Cheng, Huanying Helen Gu

    Contrastive learning improves model performance by differentiating between positive and negative sample pairs. However, its application is primarily confined to classification tasks, facing challenges with complex recognition tasks such as object detection and segmentation due to its limited capacity to capture spatial relationships and fine-grained features. To address this limitation, we propose LossTransform, a novel approach that redefines positive sample pairs and establishes a novel contrastive loss paradigm. LossTransform advances contrastive learning to the instance level, departing from the traditional sample level. Empirical evaluations on ImageNet, CIFAR, and object detection benchmarks indicate that LossTransform improves accuracy by +2.73% on CIFAR, +2.52% on ImageNet, and up to +5.2% in average precision on detection tasks, while maintaining efficiency. These results illustrate that LossTransform is compatible with large-scale training pipelines and exhibits robust performance across diverse and complex datasets.

  • Open Access

    Gait Study of Parkinson’s Disease Subjects Using Haptic Cues with A Motorized Walker

    by Minhua Zhang, N. Sertac Artan, Huanying Gu, Ziqian Dong, Lyudmila Burina Ganatra, Suzanna Shermon, Ely Rabin

    Gait abnormalities are one of the distinguishing symptoms of patients with Parkinson’s disease (PD) that contribute to fall risk. Our study compares the gait parameters of people with PD when they walk through a predefined course under different haptic speed cue conditions (1) without assistance, (2) pushing a conventional rolling walker, and (3) holding onto a self-navigating motorized walker under different speed cues. Six people with PD were recruited at the New York Institute of Technology College of Osteopathic Medicine to participate in this study. Spatial posture and gait data of the test subjects were collected via a VICON motion capture system. We developed a framework to process and extract gait features and applied statistical analysis on these features to examine the significance of the findings. The results showed that the motorized walker providing a robust haptic cue significantly improved gait symmetry of PD subjects. Specifically, the asymmetry index of the gait cycle time was reduced from 6.7% when walking without assistance to 0.56% and below when using a walker. Furthermore, the double support time of a gait cycle was reduced by 4.88% compared to walking without assistance.