Publications

Abstract

Credit-risk models are trained on proxy labels and deployed under temporal and segment change, yet no single transfer metric separates base-rate shift, probability-scale shift, and feature-label relationship change. We contribute a design-science artifact: a locked, multi-signal audit protocol for supervision drift in proxy-labeled credit-risk prediction. Five layers (transfer performance, an oracle-gap probe, a calibration diagnostic, feature-label stability, and a synthetic positive control), thresholds, and decision rules were locked before interpretation; a bounded reading is a designed outcome. On a public LendingClub dataset (temporal 2013 to 2016 and cross-segment transfer), ranking is stable and oracle gaps are small; the clearest temporal signal is a prevalence and probability-scale mismatch that intercept-only diagnostic recalibration largely reduces, though its cause is not identifiable from the available release. The positive control responds only to larger injected shifts; subtler drift cannot be excluded. Mapping diagnostic patterns to governance actions is conceptual guidance, not validated here.

Abstract

AI evaluation can support the wrong inference when an in-domain benchmark success does not survive distribution shift, or when the benchmark endpoint is entangled with a design factor. We study this problem in CRISPRi perturbation-effect prediction, evaluating a frozen Geneformer representation under a locked, pre-registered protocol: heads and model selection were frozen before test evaluation; the protocol required external outcome labels to remain withheld until final unblinding; and analysis-governing decisions were fixed before the evaluations they govern. In-distribution on the Virtual Cell Challenge (VCC), the frozen representation carries measurable predictive information beyond a dimension-matched random-feature control (ΔR² = +0.1645, 95% CI [+0.1375, +0.1920]), satisfying the pre-registered informativeness gate required before interpreting transfer. It then fails zero-shot transfer on both external screens (Spearman ρ = -0.139 and -0.267), lying below that control on each. Adding a predefined magnitude block improves the representation externally (Δρ = +0.032 and +0.143) but, under the frozen primary head, does not rescue transfer: both remain negative. A pre-registered, count-adjusted max-response secondary is positively associated with the outcome on both screens; we report it as correlational and secondary, not as a recovered magnitude signal. Finally, the VCC endpoint is strongly sample-size associated: a count-only linear model reaches R² = +0.4325, versus +0.2589 for the four magnitude scalars; adding those scalars to cell count improves R² by only +0.0017, so much of the aggregate-magnitude signal overlaps with cell count. This case study shows how locking the evaluation, harmonizing the measured endpoint, and separating primary from secondary evidence can change the inference supported by an AI benchmark.

Abstract

Transcriptomic foundation models are increasingly used as reusable cell and gene representations, but validating them on new data under weak supervision and distribution shift is hard: standard comparisons conflate genuine representation signal with model capacity, row-identity artifacts, gains over strong task-specific baselines, and outcome rules chosen after seeing the test set. We introduce a pre-registered, final-test-once evaluation framework that locks the outcome rule, seeds, and target-gene-grouped splits before any test data are seen, and scores each frozen representation against a strong expression baseline, a matched-capacity Gaussian control, and a within-split row-identity (shuffle) control; only the per-cell embedding-extraction step is model-specific. Applying it to three architecturally distinct models—Geneformer, scGPT, and UCE—across two external Replogle Perturb-seq datasets (RPE1 and K562), all three clear the capacity and row-identity controls by a wide margin, yet none reliably beats the expression baseline: the strongest (Geneformer) exceeds it by at most about 0.03 test R² and clears the pre-registered four-of-five-seed threshold in neither dataset, while scGPT and UCE fall below it. All three therefore land in the same pre-registered partial-replication category—a consistent cross-architecture outcome, even though the baseline-relative gap differs in sign and magnitude across models. These representations carry real structure beyond trivial controls but, under this weak magnitude label, do not transfer past a simple strong baseline; the locked framework is reusable for any frozen transcriptomic representation by swapping only the extraction step.

Abstract

Interpreting transcriptomic gene clusters is challenging in antimicrobial resistance research, where standard pathway enrichment methods often provide limited cluster-specific explanations and weak links to primary literature. BIOGEN addresses this challenge through a multi-agent reasoning framework that integrates biomedical evidence retrieval, structured interpretation, and multi-critic verification. Evaluated on a Salmonella enterica RNA-seq dataset, the framework achieved a BERTScore of 0.689, a Semantic Alignment Score of 0.715, a KEGG Functional Similarity of 0.342, and a hallucination rate of 0.000, compared with 0.100 for an LLM-only baseline. Across five datasets, BIOGEN maintained zero hallucination, highlighting its potential for reliable and evidence-grounded biomedical interpretation.

Abstract

Learning from weak, proxy, or relative supervision is common when ground-truth labels are unavailable, but model reliability under distribution shift remains challenging. This paper introduces supervision drift, defined as changes in the relationship between inputs, weak labels, and contexts. Using CRISPR-Cas13d transcriptomic perturbation experiments, we construct a controlled non-IID benchmark across cell lines and timepoints. The results show meaningful in-domain learning and partial cross-cell-line transfer, but major failures under temporal shift, highlighting the need for feature-stability diagnostics before deployment.

Abstract

The increasing complexity of modern smart power distribution systems (SPDSs) has made anomaly detection a significant challenge, as these systems generate vast amounts of heterogeneous and time-dependent data. Conventional detection methods often struggle with adaptability, generalization, and real-time decision-making, leading to high false alarm rates and inefficient fault detection. To address these challenges, this study proposes a novel deep reinforcement learning (DRL)-based framework, integrating a convolutional neural network (CNN) for hierarchical feature extraction and a recurrent neural network (RNN) for sequential pattern recognition and time-series modeling. To enhance model performance, we introduce a novel non-dominated sorting artificial bee colony (NSABC) algorithm, which fine-tunes the hyper-parameters of the CNN-RNN structure, including weights, biases, the number of layers and other critical parameters.

Abstract

The Internet of Things (IoT) has revolutionized modern communication systems by enabling seamless connectivity among low-power devices. However, the increasing demand for high-performance wireless networks necessitates advanced frameworks that optimize both energy efficiency (EE) and security. Cell-free massive multiple-input multiple-output (CF m-MIMO) has emerged as a promising solution for IoT networks, offering enhanced spectral efficiency, low-latency communication, and robust connectivity. Nevertheless, balancing EE and security in such systems remains a significant challenge due to the stringent power and computational constraints of IoT devices.

Abstract

Wireless sensor networks (WSNs) are a cornerstone of modern Internet of Things (IoT) infrastructure, enabling seamless data collection and communication for many IoT applications. However, the deployment of WSNs in remote or inaccessible locations poses significant challenges in terms of energy efficiency and secure communication. Sensor nodes, with their limited battery capacities, require innovative strategies to minimize energy consumption while maintaining robust network performance. Additionally, ensuring secure data transmission is critical for safeguarding the integrity and confidentiality of IoT systems.

Abstract

The Internet of Things (IoT) is integral to modern infrastructure, enabling connectivity among a wide range of devices from home automation to industrial control systems. With the exponential increase in data generated by these interconnected devices, robust anomaly detection mechanisms are essential. This paper presents a novel approach utilizing generative adversarial networks (GANs) for anomaly detection in IoT systems, optimized by a five-dimensional Gray wolf optimizer (5DGWO) to address the challenging hyperparameter tuning of GANs.

Abstract

The rapid advancement of Internet of Things (IoT) networks has revolutionized modern connectivity by integrating many low-power devices into various applications. As IoT networks expand, the demand for energy-efficient, batteryless devices becomes increasingly critical for sustainable future networks. These devices play a pivotal role in next-generation IoT applications by reducing the dependence on conventional batteries and enabling continuous operation through energy harvesting capabilities. However, several challenges hinder the widespread adoption of batteryless IoT devices, including the limited transmission range, constrained energy resources, and low spectral efficiency in IoT receivers. To address these limitations, reconfigurable intelligent surfaces (RISs) offer a promising solution by dynamically manipulating the wireless propagation environment to enhance signal strength and improve energy harvesting capabilities.

Abstract

In the evolving landscape of the smart grid (SG), the integration of non-organic multiple access (NOMA) technology has emerged as a pivotal strategy for enhancing spectral efficiency and energy management. However, the open nature of wireless channels in SG raises significant concerns regarding the confidentiality of critical control messages, especially when broadcasted from a neighborhood gateway (NG) to smart meters (SMs). This paper introduces a novel approach based on reinforcement learning (RL) to fortify the performance of secrecy. Motivated by the need for efficient and effective training of the fully connected layers in the RL network, we employ an improved chimp optimization algorithm (IChOA) to update the parameters of the RL.

Abstract

Hyperspectral (HS) image classification plays a crucial role in numerous areas including remote sensing (RS), agriculture, and the monitoring of the environment. Optimal band selection in HS images is crucial for improving the efficiency and accuracy of image classification. This process involves selecting the most informative spectral bands, which leads to a reduction in data volume. Focusing on these key bands also enhances the accuracy of classification algorithms, as redundant or irrelevant bands, which can introduce noise and lower model performance, are excluded. In this paper, we propose an approach for HS image classification using deep Q learning (DQL) and a novel multi-objective binary grey wolf optimizer (MOBGWO). We investigate the MOBGWO for optimal band selection to further enhance the accuracy of HS image classification.

Abstract

The rapid advancement of Generative AI impacted many areas in healthcare including chronic wound management. We conducted a systematic review to answer the following research question: How is generative AI used in the context of wound care and management? Our search across multiple databases resulted in more than 500 articles that matched our search criteria. After applying our inclusion/exclusion criteria, we identified 61 articles that are relevant to our research question. Preliminary analysis of these studies revealed four ways generative AI is utilized in chronic wound management context.

Abstract

The COVID-19 pandemic served as a stark reminder of a critical vulnerability in our global infrastructure: the health supply chain. The complex network responsible for delivering everything from pharmaceuticals to personal protective equipment (PPE) was strained to its breaking point, leading to widespread shortages that impacted healthcare providers and patients alike. This research examines the challenges and proposes a clear, actionable framework for strengthening this vital system through four fundamental pillars for building a resilient health supply chain.