A5082822816- Machine Learning and Data Classification
- Explainable Artificial Intelligence (XAI)
- Imbalanced Data Classification Techniques
- Cardiac, Anesthesia and Surgical Outcomes
- Hemodynamic Monitoring and Therapy
- Machine Learning in Healthcare
- Machine Learning and Algorithms
- Irish and British Studies
- Intensive Care Unit Cognitive Disorders
- Surgical Simulation and Training
- Artificial Intelligence in Healthcare and Education
- Advanced Statistical Methods and Models
- Foreign Body Medical Cases
- Data Quality and Management
- Hip and Femur Fractures
- Time Series Analysis and Forecasting
- Intestinal and Peritoneal Adhesions
- Frailty in Older Adults
- Video Analysis and Summarization
- Adversarial Robustness in Machine Learning
- Samuel Beckett and Modernism
- Transportation Planning and Optimization
- Educational Challenges and Innovations
- Markov Chains and Monte Carlo Methods
- Sparse and Compressive Sensing Techniques
Washington University in St. Louis
2020-2025
Indian Institute of Pulses Research
2022
Indian Institute of Technology Bombay
2018-2020
Telemedicine may help improve care quality and patient outcomes. for intraoperative decision support has not been rigorously studied.
Post-operative complications present a challenge to the healthcare system due high unpredictability of their incidence. Socioeconomic conditions have been established as social determinants health. However, contribution relating postoperative is still unclear it can be heterogeneous based on community, type surgical services, and sex gender. Uncovering these relations enable improved public health policy reduce such complications.
While performing Feature Subset Selection (FSS) to identify important features, a weight is assigned each feature that not necessarily meaningful or interpretable w.r.t. final task and in turn leads non-actionable information. To provide solution this problem of FSS, we introduce novel notion classification game with features as players hinge loss based characteristic function. We use the Shapley value apportion total training error explicitly compute contribution (Shapley Value Error...
An applied problem facing all areas of data science is harmonizing sources. Joining from multiple origins with unmapped and only partially overlapping features a prerequisite to developing testing robust, generalizable algorithms, especially in healthcare. This integrating usually resolved using meta-data such as feature names, which may be unavailable or ambiguous. Our goal design methods that create mapping between structured tabular datasets derived electronic health records independent...
Contrastive learning (CL) has exploded in popularity due to its ability learn effective representations using vast quantities of unlabelled data across multiple domains. CL underlies some the most impressive applications generative AI for general public. We will review fundamentals and applied work on contrastive focusing three main topics: 1) supervised, unsupervised self-supervised setup revival research as an instance discriminator. In this part, we focus about nuts bolts, such different...
Background: Anaesthesiology clinicians can implement risk mitigation strategies if they know which patients are at greatest for postoperative complications. Although machine learning models predicting complications exist, their impact on clinician assessment is unknown. Methods: This single-centre randomised clinical trial enrolled age ≥18 undergoing surgery with anaesthesiology services. providing remote intraoperative telemedicine support reviewed electronic health records (assisted group)...
With the current ongoing debate about fairness, explainability and transparency of machine learning models, their application in high-impact clinical decision-making systems must be scrutinized. We consider a real-life example risk estimation before surgery investigate potential for bias or unfairness variety algorithms. Our approach creates transparent documentation so that users can apply model carefully. augment model-card like analysis using propensity scores with decision-tree based...
Abstract Background Novel applications of telemedicine can improve care quality and patient outcomes. Telemedicine for intraoperative decision support has not been rigorously studied. Methods This single centre randomised clinical trial (RCT, clinicaltrials.gov NCT03923699 ) unselected adult surgical patients was conducted between 2019-07-01 2023-01-31. Patients received usual-care or from a service, the Anesthesiology Control Tower (ACT). The ACT provided real-time recommendations to...
We first propose an empirical risk minimization based binary classification algorithm, ExpERM, with exponential function as surrogate loss function. Our ExpERM algorithm is scalable in both the dimension of feature space well number data points it unconstrained differentiable convex optimization problem or a single constraint regularized framework. Under mild assumption on data, we show that classifier unique. implement wide collection (large-features, large-examples) real datasets. use...
An applied problem facing all areas of data science is harmonizing sources. Joining from multiple origins with unmapped and only partially overlapping features a prerequisite to developing testing robust, generalizable algorithms, especially in health care. We approach this issue the common but difficult case numeric such as nearly Gaussian binary features, where unit changes variable shift make simple matching univariate summaries unsuccessful. develop two novel procedures address problem....
In binary classification framework, we are interested in making cost sensitive label predictions the presence of uniform/symmetric noise. We first observe that $0$-$1$ Bayes classifiers not (uniform) noise robust setting. To circumvent this impossibility result, present two schemes; unlike existing methods, our schemes do require rate. The one uses $\alpha$-weighted $\gamma$-uneven margin squared loss function, $l_{\alpha, usq}$, which can handle sensitivity arising due to domain requirement...
We consider the problem of learning linear classifiers when both features and labels are binary. In addition, noisy, i.e., they could be flipped with an unknown probability. Sy-De attribute noise model, where all noisy together same probability, we show that 0-1 loss (l0−1) need not robust but a popular surrogate, squared (lsq) is. Asy-In prove l0−1 is for any distribution over 2 dimensional feature space. However, due to computational intractability l0−1, resort lsq observe it robust. Our...
We consider the problem of learning linear classifiers when both features and labels are binary. In addition, noisy, i.e., they could be flipped with an unknown probability. Sy-De attribute noise model, where all noisy together same probability, we show that $0$-$1$ loss ($l_{0-1}$) need not robust but a popular surrogate, squared ($l_{sq}$) is. Asy-In prove $l_{0-1}$ is for any distribution over 2 dimensional feature space. However, due to computational intractability $l_{0-1}$, resort...
SARS-nCoV was identified as corona virus had spread worldwide very quickly and affected more than million people worldwide. To halt this acceleration for efficient control the knowledge on genomic information is of utmost importance. We attempted to determine nature variation i.e., insertion, deletion, substitution, among structural sequences required code membrane, spike, nucleocapsid, envelope protein glycosylation between SARS CoV nCoV spike glycoproteins, respectively. Comparative...
Data missingness and quality are common problems in machine learning, especially for high-stakes applications such as healthcare. Developers often train learning models on carefully curated datasets using only high data; however, this reduces the utility of production environments. We propose a novel neural network modification to mitigate impacts low missing data which involves replacing fixed weights fully-connected layer with function an additional input. This is inspired from...
Objective: An applied problem facing all areas of data science is harmonizing sources. Joining from multiple origins with unmapped and only partially overlapping features a prerequisite to developing testing robust, generalizable algorithms, especially in health care. This joining usually resolved using meta-data, which may be unavailable or ambiguous large database. We design evaluate methods for mapping between databases independent meta-data.Methods: the challenging case numeric without...
<title>Abstract</title> <bold>Background</bold> Post-operative complications present a challenge to the healthcare system due high unpredictability of their incidence. However, socioeconomic factors that relate postoperative are still unclear as they can be heterogeneous based on communities, types surgical services, and sex gender. <bold>Methods</bold> In this study, we conducted large population cross-sectional analysis social vulnerability odds various post-surgical complications. We...