implement dynamic module selection

This work addresses three key challenges in reinforcement learning (RL): resource constraints, environmental non-stationarity, and sensitivity to random seeds—limitations inherent in static hyperparameter and architecture configurations. To this end, we propose the first theory-driven online model selection framework for RL. Our approach formulates model selection as a meta-RL problem and introduces a differentiable online selection mechanism that jointly optimizes neural architectures, learning rates, and self-model selection policies during training. Theoretical analysis establishes convergence and stability guarantees for the proposed selection criterion. Empirical evaluation across multiple standard RL benchmarks demonstrates that our method improves training efficiency by 23–41% over strong baselines, reduces policy performance variance by 57%, and maintains robust adaptation under non-stationary dynamics. These results substantiate the practical value of theoretically grounded guidance in adaptive model selection for RL.

Conventional static core-set selection fails to adapt to the heterogeneous requirements across different training stages. Method: This paper proposes a dynamic multi-objective adaptive core-set selection framework that dynamically switches sampling strategies according to training progression—emphasizing class balance in early stages, feature diversity in mid-stages, and prediction uncertainty in late stages—thereby enabling the first training-process-aware, multi-objective co-optimization. Contribution/Results: We theoretically establish a (1−1/e)-approximation guarantee. By integrating submodular optimization, active learning, and representation analysis, our method achieves O(n log n) computational efficiency. Empirically, it attains full-dataset accuracy on multiple benchmarks while significantly reducing memory overhead. Moreover, it is the first work to quantitatively characterize the dynamic evolution of data utility throughout training.

Machine learning model selection lacks formalized methodologies, making it difficult to systematically characterize contextual factors—such as data characteristics and prediction tasks—and their interactions, resulting in opaque, non-adaptive decisions. This paper introduces, for the first time, software product line (SPL) principles into ML model selection, proposing a variability-aware algorithm selection framework. It constructs a configurable feature model that explicitly captures commonalities and variabilities among contextual factors—including dataset size, feature dimensionality, and task type—as well as their logical dependencies. By integrating scikit-learn’s heuristic rules with an instantiation framework, the approach enables interpretable, adaptive, and transparent model recommendations. An empirical case study demonstrates that the method significantly outperforms existing strategies in accuracy, interpretability, and contextual adaptability.

This paper addresses the challenge of multi-model-class evaluation by proposing the Model Class Selection (MCS) framework, which identifies the collection of model classes each containing at least one optimal model—thereby enabling formal comparison of performance equivalence across model classes of differing complexity (e.g., interpretable vs. black-box models). MCS generalizes conventional model selection and Model Set Selection (MSS) by integrating likelihood maximization and risk minimization criteria via a data-splitting strategy under mild assumptions. Theoretical analysis establishes its statistical validity. Empirical evaluation—including simulations and real-data experiments—demonstrates that MCS robustly identifies simple, interpretable model classes whose predictive performance matches that of complex models. By bridging interpretability and performance assessment, MCS introduces a novel paradigm and practical tool for explainable AI research.

This work addresses the lack of structured, verifiable, and reusable decision mechanisms in existing automated machine learning approaches for model selection. It proposes a semantic task profiling–based structured agent framework that leverages retrieval-augmented generation of historical cases and code modules to construct an intermediate representation blueprint encompassing modeling components, composition logic, and execution constraints. By integrating code execution feedback with a failure-aware reinforcement learning strategy, the framework enables memory-driven, traceable, multi-stage search optimization. Evaluated on financial time-series forecasting and generation tasks, the method significantly outperforms both conventional AutoML systems and current agent-based baselines, achieving consistent improvements in task performance, execution success rate, and decision interpretability.

Existing algorithm selection models exhibit limited generalization capabilities in real-world optimization scenarios, struggling to maintain consistent performance across diverse domains. This work presents the first systematic evaluation of cross-domain generalization between synthetic benchmarks (BBOB, CEC) and practical applications—specifically robotic trajectory optimization and UAV path planning—using an algorithm selection framework grounded in problem features and historical performance data, complemented by a carefully designed cross-benchmark experimental protocol. The study uncovers the failure mechanisms and success boundaries of current approaches when deployed in realistic settings, thereby providing crucial empirical insights for developing more robust and universally applicable algorithm selection systems.

This study addresses the challenge of automatically selecting the best solver for each instance in continuous black-box optimization. It proposes a novel image-based algorithm selection approach that, for the first time, uses contour plots of objective functions as input to convolutional neural networks (CNNs), enabling direct learning of spatial landscape structures from visual representations without relying on handcrafted numerical features such as those from Exploratory Landscape Analysis (ELA). The method employs multi-view stacked or encoded contour images and is trained and evaluated within the BBOB benchmark suite and the DeepELA bi-objective evaluation framework. Experimental results demonstrate that the proposed approach significantly outperforms the single best solver on the BBOB 2009 single-objective benchmark and achieves performance comparable to feature-based methods. Further bi-objective experiments confirm its competitiveness, highlighting the effectiveness and novelty of image-driven strategies in algorithm selection.

This work addresses the challenge of feature selection in industrial settings where labeled data are scarce and multiple business constraints must be satisfied, a scenario in which conventional methods struggle to balance predictive accuracy with regulatory compliance. To this end, we propose the Model Feature Agent (MoFA) framework, which pioneers the integration of large language models’ reasoning capabilities into industrial-scale feature selection. MoFA employs structured prompts to jointly incorporate feature semantics, quantitative metrics, and domain-specific constraints, enabling interpretable and sequential automated feature engineering. Evaluations across three real-world industrial applications demonstrate that MoFA not only enhances model accuracy but also effectively uncovers high-order interaction features, yielding substantial online performance gains while producing compact, efficient, and compliant feature subsets.