2786 papers
This paper introduces Learn From the Best Evolution Strategy (LFB-ES), a novel black-box optimizer that enhances high-dimensional neural network training by iteratively guiding the population to learn from elite individuals, thereby achieving superior convergence and accuracy compared to classical methods like OpenAI-ES and CMA-ES.
This paper proposes Ephemeral Subgraph Generation (ESG), a real-time, LLM-guided approach that constructs temporary, question-specific knowledge graphs across heterogeneous engineering systems to outperform static retrieval and fixed-hop baselines in cross-system investigations, while also documenting and resolving specific software defects identified during evaluation.
This paper introduces N-SCORE, a non-work conserving, stateless core fair queuing scheduler that preserves the optimal end-to-end latency bounds of existing stateful and work-conserving techniques while significantly improving jitter suppression and scalability for deterministic networking.
This study demonstrates that while frozen DINOv2-small does not significantly outperform supervised ImageNet-pretrained ResNet50 on fully annotated panoramic radiographs, it provides superior sample efficiency and performance for tooth-level multi-label diagnosis classification under constrained annotation budgets.
This study introduces Quantum AIO-ChameleonGAN, a hybrid quantum-classical intrusion detection framework that integrates Angle of Incidence Optimization for contextual behavioural modelling with an eight-qubit Variational Quantum Circuit, achieving 99.85% accuracy and a 33.3% reduction in false negatives on the CIC-UNSW-NB15 dataset while demonstrating representational benefits through state-vector simulation.
This paper introduces the CRPC-6 dataset and demonstrates through deep learning and explainability analysis that the regional styles of Chinese paper-cutting are primarily defined by distinct craft techniques rather than thematic semantics, achieving 95% classification accuracy with a ConvNeXt-Tiny model.
This paper proposes the Decoupled Thinking–Learning–Action (D-TLA) framework, which enables autonomous robots to continuously adapt through independent yet interacting modules that maintain routine performance while significantly improving rare-event recognition and novel-task success.
This paper proposes an evolving background activity framework for lifelong autonomous robots that dynamically consolidates repeated reasoning into lightweight background processes triggered by specific conditions, thereby significantly reducing computational costs while maintaining critical event recall and improving long-horizon task success.
This paper proposes an open-ended cognitive evolution framework that enables autonomous robots to dynamically adapt their thinking, learning, and interaction mechanisms in response to unseen situations and environmental changes, significantly outperforming fixed-structure approaches in task success, safety, and efficiency.
This paper proposes an unsupervised satellite image change detection framework that reformulates a context-aware Markov Random Field with weighted 8-connected neighborhoods as a QUBO problem, achieving competitive performance on the OSCD benchmark while maintaining compatibility with future quantum annealing architectures.