4157 papers
This paper proposes Attention-Guided Stroke Location Inversion (AG-SLI), a refined, rank-based optimization framework that overcomes the instability of continuous attention-weighted gradients to significantly improve convergence speed, spatial recovery, and computational efficiency in explainable sketch recognition.
This study demonstrates that augmenting RGB inputs with hand-crafted frequency and noise-residual features fails to improve the generalization or robustness of AI-generated image detectors against unseen generators and image manipulations, often degrading performance while increasing model complexity.
This survey introduces a novel taxonomy for composing lightweight deep learning techniques—such as pruning, quantization, and knowledge distillation—to analyze how their selection, ordering, and interaction impact efficiency and accuracy, thereby offering practical guidance for deploying deep neural networks on resource-constrained edge devices.
This paper proposes a Legal Actuator Layer (LAL) that integrates Active Inference to enable autonomous drones to dynamically balance and enforce conflicting legal obligations in real time, outperforming rigid rule-based systems in complex scenarios while maintaining computational efficiency and addressing fundamental limits in learning compliance-relevant data.
This paper introduces POST, a highly efficient meta-heuristic that serves as a general post-optimizing procedure to improve the final solutions of multi-start algorithms with negligible additional computational cost, successfully enhancing the best-known results for all 48 tested instances of the planar p-median problem.
This paper demonstrates that random splitting of augmented image datasets in deep learning for surface and microscopy imaging leads to severe "sibling leakage" that artificially inflates performance metrics, and proposes a reporting protocol and a leakage statistic to ensure evaluations reflect true generalization rather than mere recognition of near-duplicates.
This study presents a mobile-centric framework for real-time flower recognition that optimizes deep learning architectures and a hybrid MobileNetV2–XGBoost pipeline for the Oxford 102 Flowers dataset, demonstrating that a tailored MobileNetV2 model converted to TensorFlow Lite achieves 96% accuracy with low latency suitable for offline Android deployment.
This study demonstrates that while the Attention-Preservation Score (APS) correlates with detection gains for specific reference detectors, it fails to serve as a reliable, generalizable guide for selecting denoising methods across diverse frozen pedestrian detectors under Gaussian noise.
This paper introduces TensorEQC, a multi-algorithmic verification framework that leverages GF(2) algebra, exact tensor networks, and reset analysis to demonstrate that effective equivalence checking requires structure-dependent engine selection rather than relying on a single universal approach.
This paper introduces FinLedger AI, a MERN-stack-based web application that integrates a SHA-256 secured blockchain for tamper-evident financial record-keeping, role-based access control, and Google Gemini-powered LLM analysis to provide proactive budget alerts and automated financial insights.