Wind energy is an important renewable energy source, and artificial intelligence (AI) plays an important role in improving its efficiency, reliability and cost-effectiveness while minimizing its environmental impact. Based on an analysis of the latest scientific literature, this article examines AI applications for the entire life cycle of wind turbines, including planning, operation and decommissioning. A key focus is on AI-driven maintenance, which reduces downtime, improves reliability and extends the lifetime of the turbines. AI also optimizes the design of wind turbines, particularly in the development of aerodynamically efficient blade shapes through rapid design iterations. In addition, AI helps to reduce the impact on the environment, e.g., by reducing bird collisions, and improves wind energy forecasting, which is essential for balancing energy flows in power systems. Despite its benefits, AI applications face challenges, including algorithmic errors, data accuracy, ethical concerns and cybersecurity risks. Further testing and validation of AI algorithms is needed to ensure their effectiveness in advancing wind energy systems.

Predicting innovation outcomes at the firm level continues to be an important but challenging goal for researchers and practitioners alike. In this study, multiple machine learning models, encompassing both ensemble-based and single-model approaches, were applied to data from the Community Innovation Survey. Methods included random forests, gradient boosting frameworks, support vector machines, neural networks, and logistic regression, each with hyperparameters optimized through Bayesian search routines and evaluated using corrected cross-validation techniques. The results showed that tree-based boosting algorithms consistently outperformed other models in accuracy, precision, F1-score, and ROC-AUC, while the kernel-based approach excelled in recall. Logistic regression proved to be the most computationally efficient model despite its weaker predictive power. The statistical analyses made it clear that the choice of an appropriate cross-validation protocol and accounting for overlapping data splits are crucial to reduce bias and ensure reliable comparisons. Overall, the results indicate that ensemble methods generally provide robust classification performance for innovation prediction tasks. However, individual models may still prove advantageous under certain metric-specific conditions or computational constraints. These observations emphasize the need to match model selection with data structure, performance objectives, and practical resource constraints when predicting and improving innovation outcomes at the firm level.

Accurate multi-class segmentation of the prostate in T2-weighted magnetic resonance imaging (MRI) into the peripheral zone (PZ), central gland (CG) and tumour is essential for targeted biopsy guidance and treatment planning. We present DSBANet, an encoder–decoder architecture that combines a pretrained ResNet-50 encoder, Atrous Spatial Pyramid Pooling, Multi-Scale Attention Fusion on skip connections, a Feature Fusion Module, deep supervision and boundary refinement. We evaluate eight architectures across three input dimensionalities (2D, 2.5D, 3D), yielding 24 models trained under identical conditions on the Prostate158 dataset. DSBANet achieves the best anatomy segmentation with PZ DSC of 0.8176 and CG DSC of 0.7888 among 2D models. To address the severe class imbalance of the tumour class, we further train DSBANet 2D with a class-weighted cross-entropy term and tumour-positive slice oversampling, raising per-case tumour DSC from 0.003 to 0.170 (a sixty-fold absolute improvement). A systematic eight-variant ablation study, evaluated under matched-pairs effect-size analysis, identifies the SE-Residual blocks and skip-connection attention as the largest contributors to tumour segmentation, while every architectural component contributes a directionally consistent gain.

This paper presents a systematic empirical evaluation of Retrieval-Augmented Generation (RAG) across 22 open-weight language models spanning five families—Qwen3.5, Llama 3.x, Gemma 3, Phi-4, and DeepSeek R1 distilled—on a domain-specific Croatian accounting benchmark derived from the publicly available mariozupan/rrif dataset. All models were deployed locally via Ollama on a single GPU server, reflecting privacy-preserving enterprise conditions. Two retrieval corpora were evaluated: a structured English Q&A-derived corpus (Corpus A, 522 accounts) and a Croatian corpus parsed from the RRiF chart of accounts (Corpus B, 2462 entries), the latter indexed under three embedding models, yielding five conditions and 116,600 queries in total. Four findings are reported. First, on Corpus A, RAG raises lookup Exact Match from near-zero to 94.9–96.2% with 18 of the 22 models converging into that band, with equivalence testing confirming practical equivalence for 87% of within-band pairs at a two-point margin; the four exceptions lie at 1B–3B parameters, but the boundary tracks instruction-following quality rather than scale. Because Corpus A shares its source with the evaluation questions, these results are an upper bound rather than an estimate of field performance. Second, embedding model choice dominates language model choice on the Croatian corpus, where a multilingual encoder raises average Exact Match by approximately 43 percentage points. Third, reasoning-distilled models show no consistent advantage on this task. Fourth, decomposing responses into mutually exclusive outcome categories shows that retrieval reduces the invention of account codes from 33.6% to below 0.3% under every condition tested, while residual error consists almost entirely of wrong-but-existing accounts whose source shifts from retrieval to selection as retrieval improves—a distinction that a binary hallucination measure conflates. Non-generative baselines outperform every model on Corpus A and ten of the 22 on Corpus B, indicating that a language model is warranted only conditionally for this task.