arXiv subjects

Vishwesh Ramanathan

Publications and source records attributed to Vishwesh Ramanathan.

3 recordsLinked to original sources

arXiv2026

DISPCA : A hybrid iterative-sequential approach for the identification of errors-in-variables model of linear DAE systems

The dynamic behavior of numerous engineering processes is effectively characterized through differential-algebraic equations (DAEs), commonly referred to as descriptor systems. While substantial progress has been achieved in identifying dynamic models governed by ordinary differential equations (ODEs), limited research has addressed the identification of descriptor systems from measured data. This work presents a systematic methodology for identifying the DAE model of a linear descriptor system in discrete difference equation form under errors-in-variables (EIV) setting, where both input and output measurements are corrupted by random noise. The proposed methodology generalizes the identification framework to handle scenarios where the system contains multiple algebraic and different ordered differential relations. The key innovation involves a partial stacking procedure of lagged data matrix with a sequentially increasing lag window that identifies all the differential relations individually. This is preceded by an iterative estimation of the measurement error covariance matrix that is diagonal and heteroskedastic, under large sample conditions. The algorithm simultaneously estimates the number of differential and algebraic relations, observability indices and delay parameters of the differential equations, and all the model coefficients directly from measured data without requiring prior specification from the user. The framework addresses the increased complexity arising from multiple dynamic coupled interactions while maintaining computational tractability through systematic decomposition of the identification problem. Effectiveness of the proposed methodology is demonstrated through several simulation studies.

eess.SY↗

arXiv2025

ModalTune: Fine-Tuning Slide-Level Foundation Models with Multi-Modal Information for Multi-task Learning in Digital Pathology

Prediction tasks in digital pathology are challenging due to the massive size of whole-slide images (WSIs) and the weak nature of training signals. Advances in computing, data availability, and self-supervised learning (SSL) have paved the way for slide-level foundation models (SLFMs) that can improve prediction tasks in low-data regimes. However, current methods under-utilize shared information between tasks and modalities. To overcome this challenge, we propose ModalTune, a novel fine-tuning framework which introduces the Modal Adapter to integrate new modalities without modifying SLFM weights. Additionally, we use large-language models (LLMs) to encode labels as text, capturing semantic relationships across multiple tasks and cancer types in a single training recipe. ModalTune achieves state-of-the-art (SOTA) results against both uni-modal and multi-modal models across four cancer types, jointly improving survival and cancer subtype prediction while remaining competitive in pan-cancer settings. Additionally, we show ModalTune is generalizable to two out-of-distribution (OOD) datasets. To our knowledge, this is the first unified fine-tuning framework for multi-modal, multi-task, and pan-cancer modeling in digital pathology.

eess.IV↗

arXiv2024

Cross-Validation Is All You Need: A Statistical Approach To Label Noise Estimation

Machine learning models experience deteriorated performance when trained in the presence of noisy labels. This is particularly problematic for medical tasks, such as survival prediction, which typically face high label noise complexity with few clear-cut solutions. Inspired by the large fluctuations across folds in the cross-validation performance of survival analyses, we design Monte-Carlo experiments to show that such fluctuation could be caused by label noise. We propose two novel and straightforward label noise detection algorithms that effectively identify noisy examples by pinpointing the samples that more frequently contribute to inferior cross-validation results. We first introduce Repeated Cross-Validation (ReCoV), a parameter-free label noise detection algorithm that is robust to model choice. We further develop fastReCoV, a less robust but more tractable and efficient variant of ReCoV suitable for deep learning applications. Through extensive experiments, we show that ReCoV and fastReCoV achieve state-of-the-art label noise detection performance in a wide range of modalities, models and tasks, including survival analysis, which has yet to be addressed in the literature. Our code and data are publicly available at https://github.com/GJiananChen/ReCoV.

cs.LG↗