The paper introduces a deep‐learning model fine‐tuned for detecting authoritarian discourse in political speeches. Set up as a regression problem with weak supervision logic, the model is trained for the task of classification of segments of text for being/not being associated with authoritarian discourse. Rather than trying to define what an authoritarian discourse is, the model builds on the assumption that authoritarian leaders inherently define it. In other words, authoritarian leaders talk like authoritarians. When combined with the discourse defined by democratic leaders, the model learns the instances that are more often associated with authoritarians on the one hand and democrats on the other. The paper discusses several evaluation tests using the model and advocates for its usefulness in a broad range of research problems. It presents a new methodology for studying latent political concepts and positions as an alternative to more traditional research strategies.

Pater's (2019) target article builds a persuasive case for establishing stronger ties between theoretical linguistics and connectionism (deep learning). This commentary extends his arguments to semantics, focusing in particular on issues of learning, compositionality, and lexical meaning.

Joe Pater's (2019) target article calls for greater interaction between neural network research and linguistics. I expand on this call and show how such interaction can benefit both fields. Linguists can contribute to research on neural networks for language technologies by clearly delineating the linguistic capabilities that can be expected of such systems, and by constructing controlled experimental paradigms that can determine whether those desiderata have been met. In the other direction, neural networks can benefit the scientific study of language by providing infrastructure for modeling human sentence processing and for evaluating the necessity of particular innate constraints on language acquisition.

The birthdate of both generative linguistics and neural networks can be taken as 1957, the year of the publication of foundational work by both Noam Chomsky and Frank Rosenblatt. This article traces the development of these two approaches to cognitive science, from their largely autonomous early development in the first thirty years, through their collision in the 1980s around the past-tense debate (Rumelhart & McClelland 1986, Pinker & Prince 1988) and their integration in much subsequent work up to the present. Although this integration has produced a considerable body of results, the continued general gulf between these two lines of research is likely impeding progress in both: on learning in generative linguistics, and on the representation of language in neural modeling. The article concludes with a brief argument that generative linguistics is unlikely to fulfill its promise of accounting for language learning if it continues to maintain its distance from neural and statistical approaches to learning.

This research paper addresses the hypothesis that sequence-based long short-term memory (LSTM) architectures improve the prediction of the next DO (days open) relative to a feed-forward multi-layer perceptron and a Cox model under strictly temporally valid predictors. Modern dairy farming can heavily benefit from optimising ‘days open’ for profitability and animal welfare. Machine learning can forecast this metric, improving farm management, disease prevention and culling decisions. This study used a dataset of 16,472 breeding records. The study compared the performance of feed-forward neural networks and two types of recurrent neural networks (RNNs). The results showed that LSTM most accurately forecasted the next ‘days open’. This demonstrates that RNN models, due to their ability to capture temporal patterns in the data, significantly outperform feed-forward and traditional statistical methods in terms of mean absolute error and concordance.

We present a critical survey on the consistency of uncertainty quantification used in deep learning and highlight partial uncertainty coverage and many inconsistencies. We then provide a comprehensive and statistically consistent framework for uncertainty quantification in deep learning that accounts for all major sources of uncertainty: input data, training and testing data, neural network weights, and machine-learning model imperfections, targeting regression problems. We systematically quantify each source by applying Bayes’ theorem and conditional probability densities and introduce a fast, practical implementation method. We demonstrate its effectiveness on a simple regression problem and a real-world application: predicting cloud autoconversion rates using a neural network trained on aircraft measurements from the Azores and guided by a two-moment bin model of the stochastic collection equation. In this application, uncertainty from the training and testing data dominates, followed by input data, neural network model, and weight variability. Finally, we highlight the practical advantages of this methodology, showing that explicitly modeling training data uncertainty improves robustness to new inputs that fall outside the training data, and enhances model reliability in real-world scenarios.

Fine-grained mortality forecasting has gained momentum in actuarial research due to its ability to capture localized, short-term fluctuations in death rates. This paper introduces MortFCNet, a deep-learning method that predicts weekly death rates using region-specific weather inputs. Unlike traditional Serfling-based methods and gradient-boosting models that rely on predefined fixed Fourier terms and manual feature engineering, MortFCNet automatically learns patterns from raw time-series data without needing explicitly defined Fourier terms or manual feature engineering. Extensive experiments across over 200 NUTS-3 regions in France, Italy, and Switzerland demonstrate that MortFCNet consistently outperforms both a standard Serfling-type baseline and XGBoost in terms of predictive accuracy. Our ablation studies further confirm its ability to uncover complex relationships in the data without feature engineering. Moreover, this work underscores a new perspective on exploring deep learning for advancing fine-grained mortality forecasting.

In deep learning, interval neural networks are used to quantify the uncertainty of a pre-trained neural network. Suppose we are given a computational problem $P$ and a pre-trained neural network $\Phi _P$ that aims to solve $P$. An interval neural network is then a pair of neural networks $(\underline {\phi }, \overline {\phi })$, with the property that $\underline {\phi }(y) \leq \Phi _P(y) \leq \overline {\phi }(y)$ for all inputs $y$, where the inequalities are meant componentwise. $(\underline {\phi }, \overline {\phi })$ are specifically trained to quantify the uncertainty of $\Phi _P$, in the sense that the size of the interval $[\underline {\phi }(y),\overline {\phi }(y)]$ quantifies the uncertainty of the prediction $\Phi _P(y)$. In this paper, we investigate the phenomenon when algorithms cannot compute interval neural networks in the setting of inverse problems. We show that in the typical setting of a linear inverse problem, the problem of constructing an optimal pair of interval neural networks is non-computable, even with the assumption that the pre-trained neural network $\Phi _P$ is an optimal solution. In other words, there exist classes of training sets $\Omega$, such that there is no algorithm, even randomised (with probability $p \geq 1/2$), that computes an optimal pair of interval neural networks for each training set ${\mathcal{T}} \in \Omega$. This phenomenon happens even when we are given a pre-trained neural network $\Phi _{{\mathcal{T}}}$ that is optimal for $\mathcal{T}$. This phenomenon is intimately linked to instability in deep learning.

Advances in deep learning and representation learning have transformed item factor analysis (IFA) in the item response theory (IRT) literature by enabling more efficient and accurate parameter estimation. Variational autoencoders (VAEs) are widely used to model high-dimensional latent variables in this context, but the limited expressiveness of their inference networks can still hinder performance. We introduce adversarial variational Bayes (AVB) and an importance-weighted extension (IWAVB) as more flexible inference algorithms for IFA. By combining VAEs with generative adversarial networks (GANs), AVB uses an auxiliary discriminator network to frame estimation as a two-player game and removes the restrictive standard normal assumption on the latent variables. Theoretically, AVB and IWAVB can achieve likelihoods that match or exceed those of VAEs and importance-weighted autoencoders (IWAEs). In exploratory analyses of empirical data, IWAVB attained higher likelihoods than IWAE, indicating greater expressiveness. In confirmatory simulations, IWAVB achieved comparable mean-square error in parameter recovery while consistently yielding higher likelihoods, and it clearly outperformed IWAE when the latent distribution was multimodal. These findings suggest that IWAVB can scale IFA to complex, large-scale, and potentially multimodal settings, supporting closer integration of psychometrics with modern multimodal data analysis.

Extreme precipitation events are projected to increase both in frequency and intensity due to climate change. High-resolution climate projections are essential to effectively model the convective phenomena responsible for severe precipitation and to plan any adaptation and mitigation action. Existing numerical methods struggle with either insufficient accuracy in capturing the evolution of convective dynamical systems, due to the low resolution, or are limited by the excessive computational demands required to achieve kilometre-scale resolution. To fill this gap, we propose a novel deep learning regional climate model (RCM) emulator called graph neural networks for climate downscaling (GNN4CD) to estimate high-resolution precipitation. The emulator is innovative in architecture and training strategy, using graph neural networks (GNNs) to learn the downscaling function through a novel hybrid imperfect framework. GNN4CD is initially trained to perform reanalysis to observation downscaling and then used for RCM emulation during the inference phase. The emulator is able to estimate precipitation at very high resolution both in space ($ 3 $km) and time ($ 1 $h), starting from lower-resolution atmospheric data ($ \sim 25 $km). Leveraging the flexibility of GNNs, we tested its spatial transferability in regions unseen during training. The model trained on northern Italy effectively reproduces the precipitation distribution, seasonal diurnal cycles, and spatial patterns of extreme percentiles across all of Italy. When used as an RCM emulator for the historical, mid-century, and end-of-century time slices, GNN4CD shows the remarkable ability to capture the shifts in precipitation distribution, especially in the tail, where changes are most pronounced.

Vibration control in structures is essential to mitigate undesired dynamic responses, thereby enhancing stability, safety, and performance under varying loading conditions. Mechanical metamaterials have emerged as effective solutions, enabling tailored dynamic properties for vibration attenuation. This study introduces a convolutional autoencoder framework for the inverse design of local resonators embedded in mechanical metamaterials. The model learns from the dynamic behaviour of primary structures coupled with ideal absorbers to predict the geometric parameters of resonators that achieve desired vibration control performance. Unlike conventional approaches requiring full numerical models, the proposed method operates as a data-driven tool, where the target frequency to be mitigated is provided as input, and the model directly outputs the resonator geometry. A large dataset, generated through physics-informed simulations of ideal absorber dynamics, supports training while incorporating both spectral and geometric variability. Within the architecture, the encoder maps input receptance spectra to resonator geometries, while the decoder reconstructs the target receptance response, ensuring dynamic consistency. Once trained, the framework predicts resonator configurations that satisfy predefined frequency targets with high accuracy, enabling efficient design of passive controllers of the syntonized mass type. This study specifically demonstrates the application of the methodology to resonators embedded in wind turbine metastructures, a critical context for mitigating structural vibrations and improving operational efficiency. Results confirm strong agreement between predicted and target responses, underscoring the potential of deep learning techniques to support on-demand inverse design of mechanical metamaterials for smart vibration control in wind energy and related engineering applications.