Search

Decoding and expressing brain activity in a comprehensible form is a challenging frontier in AI. This paper presents Thought2Text, which uses instruction-tuned Large Language Models (LLMs) fine-tuned with EEG data to achieve this goal. The approach involves three stages: (1) training an EEG encoder for visual feature extraction, (2) fine-tuning LLMs on image and text data, enabling multimodal description generation, and (3) further fine-tuning on EEG embeddings to generate text directly from EEG during inference. Experiments on a public EEG dataset collected for six subjects with image stimuli demonstrate the efficacy of multimodal LLMs (LLaMa-v3, Mistral-v0.3, Qwen2.5), validated using traditional language generation evaluation metrics, GPT-4 based assessments, and evaluations by human expert. This approach marks a significant advancement towards portable, low-cost "thoughts-to-text" technology with potential applications in both neuroscience and natural language processing (NLP).

The radiance of nighttime artificial lights measured by the VIIRS-DNB instrument on board the satellite Suomi-NPP increases at an average rate ~2.2 %/yr worldwide, whereas the artificial radiance of the night sky deduced from the Globe at Night (GAN) unaided-eye observations of the number of visible stars is reported to increase at an average rate ~9.6 %/yr. The difference between these two estimates is remarkable. This raises the question of whether the diverging temporal evolution of these indicators could be due to changes in the spectral composition of outdoor artificial light, consequence of the current process of replacement of lighting technologies. This paper presents a model for evaluating the temporal rate of change of different light pollution indicators and applies it to the VIIRS-DNB vs GAN issue, based on available data. The results show that the reported difference could be explained by spectral changes alone, if the visual observations are made under scotopic adaptation in some specific transition conditions. In case of photopic adapted observers, however, reconciling these two measurement sets requires the existence of additional light sources that affect the Globe at Night observations but do not show up in the VIIRS-DNB data. The lumen emissions of these specific sources should increase at a rate 6 %/yr worlwide, adding to the estimated 3 %/yr of the remaining lights deduced from the VIIRS-DNB measurements corrected for spectral shift., Comment: 17 pages, 1 figure. Submitted for publication

The melting temperature is important for materials design because of its relationship with thermal stability, synthesis, and processing conditions. Current empirical and computational melting point estimation techniques are limited in scope, computational feasibility, or interpretability. We report the development of a machine learning methodology for predicting melting temperatures of binary ionic solid materials. We evaluated different machine-learning models trained on a data set of the melting points of 476 non-metallic crystalline binary compounds, using materials embeddings constructed from elemental properties and density-functional theory calculations as model inputs. A direct supervised-learning approach yields a mean absolute error of around 180~K but suffers from low interpretability. We find that the fidelity of predictions can further be improved by introducing an additional unsupervised-learning step that first classifies the materials before the melting-point regression. Not only does this two-step model exhibit improved accuracy, but the approach also provides a level of interpretability with insights into feature importance and different types of melting that depend on the specific atomic bonding inside a material. Motivated by this finding, we used a symbolic learning approach to find interpretable physical models for the melting temperature, which recovered the best-performing features from both prior models and provided additional interpretability., Comment: 21 pages and 3 figures for the main manuscript, and 11 pages and 15 figures of supplementary information

Large-scale first principles molecular dynamics are crucial for simulating complex processes in chemical, biomedical, and materials sciences. However, the unfavorable time complexity with respect to system sizes leads to prohibitive computational costs when the simulation contains over a few hundred atoms in practice. We present an In-Distribution substructure Embedding Active Learner (IDEAL) to enable efficient simulation of large complex systems with quantum accuracy by maintaining a machine learning force field (MLFF) as an accurate surrogate to the first principles methods. By extracting high-uncertainty substructures into low-uncertainty atom environments, the active learner is allowed to concentrate on and learn from small substructures of interest rather than carrying out intractable quantum chemical computations on large structures. IDEAL is benchmarked on various systems and shows sub-linear complexity, accelerating the simulation thousands of times compared with conventional active learning and millions of times compared with pure first principles simulations. To demonstrate the capability of IDEAL in practical applications, we simulated a polycrystalline lithium system composed of one million atoms and the full ammonia formation process in a Haber-Bosch reaction on a 3-nm Iridium nanoparticle catalyst on a computing node comprising one single A100 GPU and 24 CPU cores.

We relate the complexity of both differential and $q$-difference equations of order one and degree one and their solutions. Our point of view is to show that if the solutions are complicated, the initial equation is complicated too. In this spirit, we bound from below an invariant of the differential or $q$-difference equation, the height of its Newton polygon, in terms of the characteristic factors of a solution. The differential and the $q$-difference cases are treated in a unified way., Comment: 27 pages, 2 figures

The Brain Imaging Data Structure (BIDS) is a community-driven standard for the organization of data and metadata from a growing range of neuroscience modalities. This paper is meant as a history of how the standard has developed and grown over time. We outline the principles behind the project, the mechanisms by which it has been extended, and some of the challenges being addressed as it evolves. We also discuss the lessons learned through the project, with the aim of enabling researchers in other domains to learn from the success of BIDS.

Background: Veterinary clinical narratives remain a largely untapped resource for addressing complex diseases. Here we compare the ability of a large language model (ChatGPT) and a previously developed regular expression (RegexT) to identify overweight body condition scores (BCS) in veterinary narratives. Methods: BCS values were extracted from 4,415 anonymised clinical narratives using either RegexT or by appending the narrative to a prompt sent to ChatGPT coercing the model to return the BCS information. Data were manually reviewed for comparison. Results: The precision of RegexT was higher (100%, 95% CI 94.81-100%) than the ChatGPT (89.3%; 95% CI82.75-93.64%). However, the recall of ChatGPT (100%. 95% CI 96.18-100%) was considerably higher than that of RegexT (72.6%, 95% CI 63.92-79.94%). Limitations: Subtle prompt engineering is needed to improve ChatGPT output. Conclusions: Large language models create diverse opportunities and, whilst complex, present an intuitive interface to information but require careful implementation to avoid unpredictable errors., Comment: Supplementary Material: The data that support the findings of this study are available in the ancillary files of this submission. 5 pages, 2 figures (textboxes)

Power electronic interfaced devices progressively enable the increasing provision of flexible operational actions in distribution networks. The feasible flexibility these devices can effectively provide requires estimation and quantification so the network operators can plan operations close to real-time. Existing approaches estimating the distribution network flexibility require the full observability of the system, meaning topological and state knowledge. However, the assumption of full observability is unrealistic and represents a barrier to system operators' adaptation. This paper proposes a definition of the distribution network flexibility problem that considers the limited observability in real-time operation. A critical review and assessment of the most prominent approaches are done based on the proposed definition. This assessment showcases the limitations and benefits of existing approaches for estimating flexibility with low observability. A case study on the CIGRE MV distribution system highlights the drawbacks brought by low observability., Comment: This paper has been accepted to the IEEE Belgrade Powertech 2023. It has 6 pages, 4 figures, and 2 tables

We study weak gravitational lensing by the cosmic large-scale structure of the 21-cm radiation background in the 3d-weak lensing formalism. The interplay between source distance measured at finite resolution, visibility and lensing terms is analysed in detail and the resulting total covariance $C_{\ell}(k,k')$ is derived. The effect of lensing correlates different multipoles through convolution, breaking the statistical homogeneity of the 21-cm radiation background. This homogeneity breaking can be exploited to reconstruct the lensing field $\hat{\phi}_{\ell m}(\kappa)$ and noise lensing reconstruction $N_{\ell}^{\hat{\phi}}$ by means of quadratic estimators. The effects related to the actual measurement process (redshift precision and visibility terms) change drastically the values of the off-diagonal terms of the total covariance $C_{\ell}(k,k')$. It is expected that the detection of lensing effects on a 21-cm radiation background will require sensitive studies and high-resolution observations by future low-frequency radio arrays such as the SKA survey., Comment: 17 pages, 4 figures, accepted by MNRAS