Praveen Rao

Ikboljon Sobirov, Cheng Xie, Muhammad Siddique et al.

Since the emergence of convolutional neural networks (CNNs) and, later, vision transformers (ViTs), deep learning architectures have predominantly relied on identical block types with varying hyperparameters. We propose a novel block alternation strategy to leverage the complementary strengths of different architectural designs, assembling structurally distinct components similar to Lego blocks. We introduce LegoNet, a deep learning framework that alternates CNN-based and SwinViT-based blocks to enhance feature learning for medical image segmentation. We investigate three variations of LegoNet and apply this concept to a previously unexplored clinical problem: the segmentation of the internal mammary artery (IMA), aorta, and perivascular adipose tissue (PVAT) from computed tomography angiography (CTA) scans. These PVAT regions have been shown to possess prognostic value in assessing cardiovascular risk and primary clinical outcomes. We evaluate LegoNet on large datasets, achieving superior performance to other leading architectures. Furthermore, we assess the model's generalizability on external testing cohorts, where an expert clinician corrects the model's segmentations, achieving DSC > 0.90 across various external, international, and public cohorts. To further validate the model's clinical reliability, we perform intra- and inter-observer variability analysis, demonstrating strong agreement with human annotations. The proposed methodology has significant implications for diagnostic cardiovascular management and early prognosis, offering a robust, automated solution for vascular and perivascular segmentation and risk assessment in clinical practice, paving the way for personalised medicine.

Praveen Rao

Large language models (LLMs) require enormous computing power to pretrain on massive datasets. When limited datasets are available, smaller-sized LLMs are better choice to pretrain (on user-specified datasets) by following the scaling laws of LLMs. Using pretrained models, vector embeddings can be generated for raw data and stored using vector databases to support modern AI applications and semantic search. In this work, we investigate the performance of pretraining techniques for smaller-sized LLMs on an experimental testbed (with commodity GPUs) available to academic users at no charge. We consider data parallelism, intra-operator parallelism, and inter-operator/pipeline parallelism, and their combinations for pretraining. We set up different GPU clusters with homogeneous and heterogeneous GPU hardware. Furthermore, we investigate the impact of network latency on pretraining performance especially when GPUs are geographically distributed. We used GPT-2 medium and large models and pretrained them using open-source packages, namely, Alpa and Ray. We observed that Alpa's execution plans that collectively optimized intra-operator and inter-operator/pipeline parallelism consistently performed the best when GPUs were geographically distributed. This was especially true when the network latencies were in 10's of milliseconds. Based on the insights gained from the experiments, we propose a systematic approach for selecting the appropriate pretraining technique to achieve high training performance/lower execution time as well as to reduce the number of GPUs used.

Shivika Prasanna, Ajay Kumar, Deepthi Rao et al.

The integration of knowledge graphs and graph machine learning (GML) in genomic data analysis offers several opportunities for understanding complex genetic relationships, especially at the RNA level. We present a comprehensive approach for leveraging these technologies to analyze genomic variants, specifically in the context of RNA sequencing (RNA-seq) data from COVID-19 patient samples. The proposed method involves extracting variant-level genetic information, annotating the data with additional metadata using SnpEff, and converting the enriched Variant Call Format (VCF) files into Resource Description Framework (RDF) triples. The resulting knowledge graph is further enhanced with patient metadata and stored in a graph database, facilitating efficient querying and indexing. We utilize the Deep Graph Library (DGL) to perform graph machine learning tasks, including node classification with GraphSAGE and Graph Convolutional Networks (GCNs). Our approach demonstrates significant utility using our proposed tool, VariantKG, in three key scenarios: enriching graphs with new VCF data, creating subgraphs based on user-defined features, and conducting graph machine learning for node classification.

Abdulkadir Korkmaz, Praveen Rao

Federated learning (FL) has come forward as a critical approach for privacy-preserving machine learning in healthcare, allowing collaborative model training across decentralized medical datasets without exchanging clients' data. However, current security implementations for these systems face a fundamental trade-off: rigorous cryptographic protections like fully homomorphic encryption (FHE) impose prohibitive computational overhead, while lightweight alternatives risk vulnerable data leakage through model updates. To address this issue, we present FAS (Fast and Secure Federated Learning), a novel approach that strategically combines selective homomorphic encryption, differential privacy, and bitwise scrambling to achieve robust security without compromising practical usability. Our approach eliminates the need for model pretraining phases while dynamically protecting high-risk model parameters through layered encryption and obfuscation. We implemented FAS using the Flower framework and evaluated it on a cluster of eleven physical machines. Our approach was up to 90\% faster than applying FHE on the model weights. In addition, we eliminated the computational overhead that is required by competitors such as FedML-HE and MaskCrypt. Our approach was up to 1.5$\times$ faster than the competitors while achieving comparable security results. Experimental evaluations on medical imaging datasets confirm that FAS maintains similar security results to conventional FHE against gradient inversion attacks while preserving diagnostic model accuracy. These results position FAS as a practical solution for latency-sensitive healthcare applications where both privacy preservation and computational efficiency are requirements.

Polycarp Nalela, Deepthi Rao, Praveen Rao

Cancer remains a leading global health challenge and a major cause of mortality. This study leverages machine learning (ML) to predict the survivability of cancer patients with metastatic patterns using the comprehensive MSK-MET dataset, which includes genomic and clinical data from 25,775 patients across 27 cancer types. We evaluated five ML models-XGBoost, Naïve Bayes, Decision Tree, Logistic Regression, and Random Fores using hyperparameter tuning and grid search. XGBoost emerged as the best performer with an area under the curve (AUC) of 0.82. To enhance model interpretability, SHapley Additive exPlanations (SHAP) were applied, revealing key predictors such as metastatic site count, tumor mutation burden, fraction of genome altered, and organ-specific metastases. Further survival analysis using Kaplan-Meier curves, Cox Proportional Hazards models, and XGBoost Survival Analysis identified significant predictors of patient outcomes, offering actionable insights for clinicians. These findings could aid in personalized prognosis and treatment planning, ultimately improving patient care.

Shivika Prasanna, Deepthi Rao, Eduardo Simoes et al.

Real-world knowledge can be represented as a graph consisting of entities and relationships between the entities. The need for efficient and scalable solutions arises when dealing with vast genomic data, like RNA-sequencing. Knowledge graphs offer a powerful approach for various tasks in such large-scale genomic data, such as analysis and inference. In this work, variant-level information extracted from the RNA-sequences of vaccine-naïve COVID-19 patients have been represented as a unified, large knowledge graph. Variant call format (VCF) files containing the variant-level information were annotated to include further information for each variant. The data records in the annotated files were then converted to Resource Description Framework (RDF) triples. Each VCF file obtained had an associated CADD scores file that contained the raw and Phred-scaled scores for each variant. An ontology was defined for the VCF and CADD scores files. Using this ontology and the extracted information, a large, scalable knowledge graph was created. Available graph storage was then leveraged to query and create datasets for further downstream tasks. We also present a case study using the knowledge graph and perform a classification task using graph machine learning. We also draw comparisons between different Graph Neural Networks (GNNs) for the case study.

Raisa M. Simoes, Timoteo Kelly, Eduardo J. Simoes et al.

This paper aims to explore two competing data science methodologies to attempt answering the question, "Which issues contributed most to voters' choice in the 2024 presidential election?" The methodologies involve novel empirical evidence driven by artificial intelligence (AI) techniques. By using two distinct methods based on natural language processing and clustering analysis to mine over eight thousand user comments on election-related YouTube videos from one right leaning journal, Wall Street Journal, and one left leaning journal, New York Times, during pre-election week, we quantify the frequency of selected issue areas among user comments to infer which issues were most salient to potential voters in the seven days preceding the November 5th election. Empirically, we primarily demonstrate that immigration and democracy were the most frequently and consistently invoked issues in user comments on the analyzed YouTube videos, followed by the issue of identity politics, while inflation was significantly less frequently referenced. These results corroborate certain findings of post-election surveys but also refute the supposed importance of inflation as an election issue. This indicates that variations on opinion mining, with their analysis of raw user data online, can be more revealing than polling and surveys for analyzing election outcomes.

Hulayyil Alshammari, Praveen Rao

Large language models (LLMs) have rapidly transformed the creation of written materials. LLMs have led to questions about writing integrity, thereby driving the creation of artificial intelligence (AI) detection technologies. Adversarial attacks, such as standard and humanized paraphrasing, inhibit detectors' ability to detect machine-generated text. Previous studies have mainly focused on ChatGPT and other well-known LLMs and have shown varying accuracy across detectors. However, there is a clear gap in the literature about DeepSeek, a recently published LLM. Therefore, in this work, we investigate whether six generally accessible AI detection tools -- AI Text Classifier, Content Detector AI, Copyleaks, QuillBot, GPT-2, and GPTZero -- can consistently recognize text generated by DeepSeek. The detectors were exposed to the aforementioned adversarial attacks. We also considered DeepSeek as a detector by performing few-shot prompting and chain-of-thought reasoning (CoT) for classifying AI and human-written text. We collected 49 human-authored question-answer pairs from before the LLM era and generated matching responses using DeepSeek-v3, producing 49 AI-generated samples. Then, we applied adversarial techniques such as paraphrasing and humanizing to add 196 more samples. These were used to challenge detector robustness and assess accuracy impact. While QuillBot and Copyleaks showed near-perfect performance on original and paraphrased DeepSeek text, others -- particularly AI Text Classifier and GPT-2 -- showed inconsistent results. The most effective attack was humanization, reducing accuracy to 71% for Copyleaks, 58% for QuillBot, and 52% for GPTZero. Few-shot and CoT prompting showed high accuracy, with the best five-shot result misclassifying only one of 49 samples (AI recall 96%, human recall 100%).

Ahmad Alhonainy, Praveen Rao

Federated Learning (FL) allows multiple distributed devices to jointly train a shared model without centralizing data, but communication cost remains a major bottleneck, especially in resource-constrained environments. This paper introduces caching strategies - FIFO, LRU, and Priority-Based - to reduce unnecessary model update transmissions. By selectively forwarding significant updates, our approach lowers bandwidth usage while maintaining model accuracy. Experiments on CIFAR-10 and medical datasets show reduced communication with minimal accuracy loss. Results confirm that intelligent caching improves scalability, memory efficiency, and supports reliable FL in edge IoT networks, making it practical for deployment in smart cities, healthcare, and other latency-sensitive applications.

Timoteo Kelly, Abdulkadir Korkmaz, Samuel Mallet et al.

We present Health App Reviews for Privacy & Trust (HARPT), a large-scale annotated corpus of user reviews from Electronic Health (eHealth) applications (apps) aimed at advancing research in user privacy and trust. The dataset comprises 480K user reviews labeled in seven categories that capture critical aspects of trust in applications (TA), trust in providers (TP), and privacy concerns (PC). Our multistage strategy integrated keyword-based filtering, iterative manual labeling with review, targeted data augmentation, and weak supervision using transformer-based classifiers. In parallel, we manually annotated a curated subset of 7,000 reviews to support the development and evaluation of machine learning models. We benchmarked a broad range of models, providing a baseline for future work. HARPT is released under an open resource license to support reproducible research in usable privacy and trust in digital libraries and health informatics.

Shraboni Sarker, Ahmad Tamim Hamad, Hulayyil Alshammari et al.

Large language models have gained tremendous popularity in domains such as e-commerce, finance, healthcare, and education. Fine-tuning is a common approach to customize an LLM on a domain-specific dataset for a desired downstream task. In this paper, we present a valuable resource for fine-tuning LLMs developed for the Spanish language to perform a variety of tasks such as classification, masked language modeling, clustering, and others. Our resource is a collection of handwritten notary records from the seventeenth century obtained from the National Archives of Argentina. This collection contains a combination of original images and transcribed text (and metadata) of 160+ pages that were handwritten by two notaries, namely, Estenban Agreda de Vergara and Nicolas de Valdivia y Brisuela nearly 400 years ago. Through empirical evaluation, we demonstrate that our collection can be used to fine-tune Spanish LLMs for tasks such as classification and masked language modeling, and can outperform pre-trained Spanish models and ChatGPT-3.5/ChatGPT-4o. Our resource will be an invaluable resource for historical text analysis and is publicly available on GitHub.

Arun Zachariah, Praveen Rao, Brian Corn et al.

The 2030 Challenge is aimed at making all new buildings and major renovations carbon neutral by 2030. One of the potential solutions to meet this challenge is through innovative sustainable design strategies. For developing such strategies it is important to understand how the various building factors contribute to energy usage of a building, right at design time. The growth of artificial intelligence (AI) in recent years provides an unprecedented opportunity to advance sustainable design by learning complex relationships between building factors from available data. However, rich training datasets are needed for AI-based solutions to achieve good prediction accuracy. Unfortunately, obtaining training datasets are time consuming and expensive in many real-world applications. Motivated by these reasons, we address the problem of accurately predicting the energy usage of new or unknown building types, i.e., those building types that do not have any training data. We propose a novel approach based on zero-shot learning (ZSL) to solve this problem. Our approach uses side information from building energy modeling experts to predict the closest building types for a given new/unknown building type. We then obtain the predicted energy usage for the k-closest building types using the models learned during training and combine the predicted values using a weighted averaging function. We evaluated our approach on a dataset containing five building types generated using BuildSimHub, a popular platform for building energy modeling. Our approach achieved better average accuracy than a regression model (based on XGBoost) trained on the entire dataset of known building types.

Mohamoud Ali, Yugyung Lee, Praveen Rao

Dynamic networks have intrinsic structural, computational, and multidisciplinary advantages. Link prediction estimates the next relationship in dynamic networks. However, in the current link prediction approaches, only bipartite or non-bipartite but homogeneous networks are considered. The use of adjacency matrix to represent dynamically evolving networks limits the ability to analytically learn from heterogeneous, sparse, or forming networks. In the case of a heterogeneous network, modeling all network states using a binary-valued matrix can be difficult. On the other hand, sparse or currently forming networks have many missing edges, which are represented as zeros, thus introducing class imbalance or noise. We propose a time-parameterized matrix (TP-matrix) and empirically demonstrate its effectiveness in non-bipartite, heterogeneous networks. In addition, we propose a predictive influence index as a measure of a node's boosting or diminishing predictive influence using backward and forward-looking maximization over the temporal space of the n-degree neighborhood. We further propose a new method of canonically representing heterogeneous time-evolving activities as a temporally parameterized network model (TPNM). The new method robustly enables activities to be represented as a form of a network, thus potentially inspiring new link prediction applications, including intelligent business process management systems and context-aware workflow engines. We evaluated our model on four datasets of different network systems. We present results that show the proposed model is more effective in capturing and retaining temporal relationships in dynamically evolving networks. We also show that our model performed better than state-of-the-art link prediction benchmark results for networks that are sensitive to temporal evolution.