Machine Learning and Knowledge Extraction

Large Language Models (LLMs) offer new opportunities to devise automated implementation generation methods that can tackle problem solving beyond traditional methods, which usually require algorithmic specifications and use only static domain knowledge. LLMs can support devising new methods to support activities in tackling open-ended problems, like problem framing, exploring possible solving approaches, feature elaboration and combination, advanced implementation assessment, and handling unexpected situations. This paper presents a detailed overview of the current work on LLMs, including model prompting, retrieval-augmented generation (RAG), and reinforcement learning. It then proposes a novel, LLM-based Cognitive Architecture (CA) to generate programming code starting from verbal discussions in natural language, a particular kind of problem-solving activity. The CA uses four strategies, three top-down and one bottom-up, to elaborate, adaptively process, memorize, and learn. Experiments are devised to study the CA performance, e.g., convergence rate, semantic fidelity, and code correctness.

Noise is one of the main pollutants in urban contexts, even if it is not perceived as severe as other pollutants. Transportation, specifically road traffic, accounts for most of the urban environmental noise, and its monitoring is very important and sometimes compelled by law. To do this, two different approaches are possible: a direct measurement campaign or a simulation approach. The so-called Road Traffic Noise Models (RTNMs) are used for this second scope. In recent years, noise assessment has also been experimented with through Machine Learning (ML) techniques: ML is very interesting mainly because it is usable in unusual road traffic conditions, like in the presence of roundabouts and/or stops and traffic lights, or more generally when the free flow aspect is not verified, and the classic RTNMs fail. In this contribution, a large and comprehensive study on four different ML regressors is presented. After careful hyperparameter tuning, regressors have been calibrated by using two different approaches: a classic train/test split on real road traffic data, and by using a computed dataset. Results show a quantitative and qualitative description of the outputs of the ML regressors functioning, and how their calibration by using computed data instead of real data can give good output simulations.

Federated learning (FL) provides a privacy-preserving approach for training machine learning models across distributed datasets; however, its deployment in environmental monitoring remains underexplored. This paper uses the WHIN dataset, comprising 144 weather stations across Indiana, to establish a benchmark for FL in soil moisture prediction. The work presents three primary contributions: the design of lightweight CNNs optimized for edge deployment, a comprehensive robustness assessment of FL under non-IID and adversarial conditions, and the development of a large-scale, reproducible agricultural FL benchmark using the WHIN network. The paper designs and evaluates lightweight (∼0.8 k parameters) and heavy (∼9.4 k parameters) convolutional neural networks (CNNs) under both centralized and federated settings, supported by ablation studies on feature importance and model architecture. Results show that lightweight CNNs achieve near-heavy CNN performance (MAE = 7.8 cbar vs. 7.6 cbar) while reducing computation and communication overhead. Beyond accuracy, this work systematically benchmarks robustness under adversarial and non-IID conditions, providing new insights for deploying federated models in agricultural IoT.