arXiv subjects

Anjishnu Banerjee

Publications and source records attributed to Anjishnu Banerjee.

9 recordsLinked to original sources

arXiv2026

Model Retirement Creates Reproducibility Risk in Biomedical AI Publications

Background. Large language models (LLMs) are being adopted in biomedical research at a rapid and accelerating pace, yet commercial services that host many widely used models operate under deprecation schedules that can complicate scientific reproducibility. Methods. We searched PubMed for original research articles from 2022 through March 2026 that applied a specific LLM to a biomedical task. An extraction agent identified model names from 61,077 article abstracts with human reviewers validating a subset for extraction accuracy. Extracted model names were normalized to canonical model identifiers. Lifecycle data (release date, retirement date, status) were compiled for the 50 most frequently used models. Results. We identified 8,931 paper-model mentions spanning 5,242 unique publications after restricting the analysis to the 50 most frequently used models. Among these mentions, 77.7% cited a commercial closed-weight model. Overall, 42% involved a model that was already retired by the time of official publication or is scheduled to retire within two years of publication. The median interval from publication to model retirement was 538 days. Conclusion. Many biomedical publications using LLMs are on a trajectory toward computational non-reproducibility after publication. Model deprecation should be treated as a core reporting and preservation issue for biomedical research.

cs.AI↗

arXiv2026

Rising Prevalence of Detected AI-Generated Text in Medical Literature: Longitudinal Analysis in Open Access Articles

Generative artificial intelligence (AI) tools are becoming increasingly used for writing tasks. However, the extent of their use in peer-reviewed medical literature remains unclear. We conducted a longitudinal analysis of all Original Investigations, Research Letters, and Invited Commentaries published in JAMA Network Open from January 2022 through March 2025. The main body text of 7,251 articles was analyzed using a commercial AI-detection tool (Originality.AI) to estimate the probability that manuscripts contained a significant amount of AI-generated content. Articles were analyzed aggregated by month, publication type, and domain. Overall, 195 articles (2.7%) were classified as containing significant AI-generated text. The monthly proportion increased from 0.0% in January 2022 to 11.3% in March 2025, with a significant upward trend over time (P<0.001). Invited Commentaries had the highest proportion of detected AI-generated content (6.7%), followed by Original Investigations (2.2%) and Research Letters (1.4%). There was also significant variation across publication domain (P=0.04). Only 15 articles (0.2%) disclosed large language model use, of which 40.0% were classified as containing AI-generated text. While findings suggest increasing detectable AI-generated content in medical literature, limitations of current detection tools necessitates cautious interpretation.

cs.DL↗

arXiv2025

A Novel Algorithm for Personalized Federated Learning: Knowledge Distillation with Weighted Combination Loss

Federated learning (FL) offers a privacy-preserving framework for distributed machine learning, enabling collaborative model training across diverse clients without centralizing sensitive data. However, statistical heterogeneity, characterized by non-independent and identically distributed (non-IID) client data, poses significant challenges, leading to model drift and poor generalization. This paper proposes a novel algorithm, pFedKD-WCL (Personalized Federated Knowledge Distillation with Weighted Combination Loss), which integrates knowledge distillation with bi-level optimization to address non-IID challenges. pFedKD-WCL leverages the current global model as a teacher to guide local models, optimizing both global convergence and local personalization efficiently. We evaluate pFedKD-WCL on the MNIST dataset and a synthetic dataset with non-IID partitioning, using multinomial logistic regression and multilayer perceptron models. Experimental results demonstrate that pFedKD-WCL outperforms state-of-the-art algorithms, including FedAvg, FedProx, Per-FedAvg, and pFedMe, in terms of accuracy and convergence speed.

stat.ML↗

arXiv2024

Stronger Baseline Models -- A Key Requirement for Aligning Machine Learning Research with Clinical Utility

Machine Learning (ML) research has increased substantially in recent years, due to the success of predictive modeling across diverse application domains. However, well-known barriers exist when attempting to deploy ML models in high-stakes, clinical settings, including lack of model transparency (or the inability to audit the inference process), large training data requirements with siloed data sources, and complicated metrics for measuring model utility. In this work, we show empirically that including stronger baseline models in healthcare ML evaluations has important downstream effects that aid practitioners in addressing these challenges. Through a series of case studies, we find that the common practice of omitting baselines or comparing against a weak baseline model (e.g. a linear model with no optimization) obscures the value of ML methods proposed in the research literature. Using these insights, we propose some best practices that will enable practitioners to more effectively study and deploy ML models in clinical settings.

cs.LG↗

arXiv2021

Nonparametric Bayesian Knockoff Generators for Feature Selection Under Complex Data Structure

The recent proliferation of high-dimensional data, such as electronic health records and genetics data, offers new opportunities to find novel predictors of outcomes. Presented with a large set of candidate features, interest often lies in selecting the ones most likely to be predictive of an outcome for further study. Controlling the false discovery rate (FDR) at a specified level is often desired in evaluating these variables. Knockoff filtering is an innovative strategy for conducting FDR-controlled feature selection. This paper proposes a nonparametric Bayesian model for generating high-quality knockoff copies that can improve the accuracy of predictive feature identification for variables arising from complex distributions, which can be skewed, highly dispersed and/or a mixture of distributions. This paper provides a detailed description for generating knockoff copies from a GDPM model via MCMC posterior sampling. Additionally, we provide a theoretical guarantee on the robustness of the knockoff procedure. Through simulations, the method is shown to identify important features with accurate FDR control and improved power over the popular second-order Gaussian knockoff generator. Furthermore, the model is compared with finite Gaussian mixture knockoff generator in FDR and power. The proposed technique is applied for detecting genes predictive of survival in ovarian cancer patients using data from The Cancer Genome Atlas (TCGA).

stat.ME↗

arXiv2019

Radiomic features of multi-parametric MRI present stable associations with analogous histological features in brain cancer patients

MR-derived radiomic features have demonstrated substantial predictive utility in modeling different prognostic factors of glioblastomas and other brain cancers. However, the biological relationship underpinning these predictive models has been largely unstudied, with the generalizability of these models also called into question. Here, we examine the localized relationship between MR-derived radiomic features and histology-derived histomic features using a dataset of 16 brain cancer patients. Tile-based radiomics features were collected on T1W, post-contrast T1W, FLAIR, and DWI-derived ADC images acquired prior to patient death, with analogous histomic features collected for autopsy samples co-registered to the MRI. Features were collected for each original image, as well as a 3D wavelet decomposition of each image, resulting in 837 features per MR image and histology image. Correlative analyses were used to assess the degree of association between radiomic-histomic pairs for each MRI. The influence of several confounds were also assessed using linear mixed effect models for the normalized radiomic-histomic distance, testing for main effects of scanners from different vendors and acquisition field strength. Results as a whole were largely heterogenous, but several features demonstrated substantial associations with their histomic analogs, particularly those derived from the FLAIR and post-contrast T1W images. These most-associated features typically presented as stable across confounding factors as well. These data suggest that a subset of radiomic features are able to consistently capture texture information about the underlying tissue histology.

q-bio.QM↗

arXiv2013

Parallel inversion of huge covariance matrices

An extremely common bottleneck encountered in statistical learning algorithms is inversion of huge covariance matrices, examples being in evaluating Gaussian likelihoods for a large number of data points. We propose general parallel algorithms for inverting positive definite matrices, which are nearly rank deficient. Such matrix inversions are needed in Gaussian process computations, among other settings, and remain a bottleneck even with the increasing literature on low rank approximations. We propose a general class of algorithms for parallelizing computations to dramatically speed up computation time by orders of magnitude exploiting multicore architectures. We implement our algorithm on a cloud computing platform, providing pseudo and actual code. The algorithm can be easily implemented on any multicore parallel computing resource. Some illustrations are provided to give a flavor for the gains and what becomes possible in freeing up this bottleneck.

stat.ME↗

arXiv2013

Bayesian learning of joint distributions of objects

There is increasing interest in broad application areas in defining flexible joint models for data having a variety of measurement scales, while also allowing data of complex types, such as functions, images and documents. We consider a general framework for nonparametric Bayes joint modeling through mixture models that incorporate dependence across data types through a joint mixing measure. The mixing measure is assigned a novel infinite tensor factorization (ITF) prior that allows flexible dependence in cluster allocation across data types. The ITF prior is formulated as a tensor product of stick-breaking processes. Focusing on a convenient special case corresponding to a Parafac factorization, we provide basic theory justifying the flexibility of the proposed prior and resulting asymptotic properties. Focusing on ITF mixtures of product kernels, we develop a new Gibbs sampling algorithm for routine implementation relying on slice sampling. The methods are compared with alternative joint mixture models based on Dirichlet processes and related approaches through simulations and real data applications.

stat.ME↗

arXiv2011

Efficient Gaussian Process Regression for Large Data Sets

Gaussian processes (GPs) are widely used in nonparametric regression, classification and spatio-temporal modeling, motivated in part by a rich literature on theoretical properties. However, a well known drawback of GPs that limits their use is the expensive computation, typically O($n^3$) in performing the necessary matrix inversions with $n$ denoting the number of data points. In large data sets, data storage and processing also lead to computational bottlenecks and numerical stability of the estimates and predicted values degrades with $n$. To address these problems, a rich variety of methods have been proposed, with recent options including predictive processes in spatial data analysis and subset of regressors in machine learning. The underlying idea in these approaches is to use a subset of the data, leading to questions of sensitivity to the subset and limitations in estimating fine scale structure in regions that are not well covered by the subset. Motivated by the literature on compressive sensing, we propose an alternative random projection of all the data points onto a lower-dimensional subspace. We demonstrate the superiority of this approach from a theoretical perspective and through the use of simulated and real data examples. Some Keywords: Bayesian; Compressive Sensing; Dimension Reduction; Gaussian Processes; Random Projections; Subset Selection

stat.ME↗