Search

Ontologies are useful for automatic machine processing of domain knowledge as they represent it in a structured format. Yet, constructing ontologies requires substantial manual effort. To automate part of this process, large language models (LLMs) have been applied to solve various subtasks of ontology learning. However, this partial ontology learning does not capture the interactions between subtasks. We address this gap by introducing OLLM, a general and scalable method for building the taxonomic backbone of an ontology from scratch. Rather than focusing on subtasks, like individual relations between entities, we model entire subcomponents of the target ontology by finetuning an LLM with a custom regulariser that reduces overfitting on high-frequency concepts. We introduce a novel suite of metrics for evaluating the quality of the generated ontology by measuring its semantic and structural similarity to the ground truth. In contrast to standard metrics, our metrics use deep learning techniques to define more robust distance measures between graphs. Both our quantitative and qualitative results on Wikipedia show that OLLM outperforms subtask composition methods, producing more semantically accurate ontologies while maintaining structural integrity. We further demonstrate that our model can be effectively adapted to new domains, like arXiv, needing only a small number of training examples. Our source code and datasets are available at https://github.com/andylolu2/ollm.

This study examines the relationship between early solar coronal mass ejection (CME) propagation, the associated filament eruption, and coronal dimming in the rare event observed on March 28, 2022, which featured a three-part CME in the low corona of active region AR 12975, including a bright core/filament, dark cavity, and bright front edge. We employ 3D filament and CME shock reconstructions using data from SolO, STEREO-A, and SDO to track the filament's path, height, and kinematics. Our analysis across three viewpoints shows the outer front in SolO/EUI 304 \r{A} aligns with shock structures in STEREO-A/EUVI 195 \r{A}, showing a full 3D EUV wave dome, later matching the outer CME front in STEREO-A COR2. We introduce the method ATLAS-3D (Advanced Technique for single Line-of-sight Acquisition of Structures in 3D) and validate it against traditional approaches to reconstruct CME shock using SOLO data exclusively. Additionally, we estimate early CME propagation characteristics based on coronal dimming evolution with the DIRECD method. Results indicate that the filament height increased from 28 to 616 Mm (0.04 to 0.89 Rs) within 30 minutes (11:05 to 11:35 UT), reaching peak velocity of around 648 km/s and acceleration of around 1624 m/s$^2$. At 11:45 UT, the filament deflected by 12{\deg} to a height of 841 Mm (1.21 Rs), while the CME shock expanded from 383 to 837 Mm (0.55 to 1.2 Rs) over 10 minutes. Key parameters include a CME direction inclined by 6{\deg}, a 21{\deg} half-width, and a 1.12 Rs cone height at the dimming's impulsive phase end. This event demonstrates that expanding dimming correlates with early CME development, with the DIRECD method linking 2D dimming to 3D CME evolution. These insights underscore the value of multi-viewpoint observations and advanced reconstructions for improving space weather forecasting., Comment: 14 pages, 9 main figures, 2 appendix figures, 1 table, accepted for Publication in Astronomy and Astrophysics

On May 8, 2024, solar active region 13664 produced an X-class flare, several M-class flares, and multiple Earth-directed Coronal Mass Ejections (CMEs). The initial CME caused coronal dimmings, characterized by localized reductions in extreme-ultraviolet (EUV) emissions, indicating mass loss and expansion during the eruption. After one solar rotation, on June 8, 2024, the same region produced another M-class flare followed by coronal dimmings observed by the SDO and STEREO spacecraft. We analyzed early CME evolution and direction from coronal dimming expansion at the end of the impulsive phase using the DIRECD (Dimming Inferred Estimation of CME Direction) method. To validate the 3D CME cone, we compared CME properties from the low corona with white-light coronagraph data. The May 8 CME expanded radially, with a 7.7 deg inclination, 70 deg angular width, and 0.81 Rsun cone height, while the June 8 CME had a 15.7 deg inclination, 81 deg width, and 0.89 Rsun height. Our study shows that tracking low coronal signatures, like coronal dimming expansion, can estimate CME direction early, providing crucial lead time for space weather forecasts., Comment: Accepted for Publication in Astronomy and Astrophysics. arXiv admin note: text overlap with arXiv:2311.13942

We analyze the interaction between an Interplanetary Coronal Mass Ejection (ICME) detected in situ at the L1 Lagrange point on 2016 October 12 with a trailing High-Speed Stream (HSS). We aim to estimate the region in the interplanetary (IP) space where the interaction happened/started using a combined observational-modeling approach. We use Minimum Variance Analysis and the Walen test to analyze possible reconnection exhaust at the interface of ICME and HSS. We perform a Graduated Cylindrical Shell reconstruction of the CME to estimate the geometry and source location of the CME. Finally, we use a two-step Drag Based Model (DBM) model to estimate the region in IP space where the interaction took place. The magnetic obstacle (MO) observed in situ shows a fairly symmetric and undisturbed structure and shows the magnetic flux, helicity, and expansion profile/speed of a typical ICME. The MVA together with the Walen test, however, confirms reconnection exhaust at the ICME HSS boundary. Thus, in situ signatures are in favor of a scenario where the interaction is fairly recent. The trailing HSS shows a distinct velocity profile which first reaches a semi-saturated plateau with an average velocity of 500 km/s and then saturates at a maximum speed of 710 km/s . We find that the HSS interaction with the ICME is influenced only by this initial plateau. The results of the two-step DBM suggest that the ICME has started interacting with the HSS close to Earth (approx 0.81 au), which compares well with the deductions from in situ signatures., Comment: 14 pages, 8 Figures, Accepted for publication in the Astrophysical Journal

Deep neural networks trained via empirical risk minimisation often exhibit significant performance disparities across groups, particularly when group and task labels are spuriously correlated (e.g., "grassy background" and "cows"). Existing bias mitigation methods that aim to address this issue often either rely on group labels for training or validation, or require an extensive hyperparameter search. Such data and computational requirements hinder the practical deployment of these methods, especially when datasets are too large to be group-annotated, computational resources are limited, and models are trained through already complex pipelines. In this paper, we propose Targeted Augmentations for Bias Mitigation (TAB), a simple hyperparameter-free framework that leverages the entire training history of a helper model to identify spurious samples, and generate a group-balanced training set from which a robust model can be trained. We show that TAB improves worst-group performance without any group information or model selection, outperforming existing methods while maintaining overall accuracy., Comment: Accepted at the 18th European Conference on Computer Vision (ECCV 2024) as an Oral presentation

Data collection is often difficult in critical fields such as medicine, physics, and chemistry. As a result, classification methods usually perform poorly with these small datasets, leading to weak predictive performance. Increasing the training set with additional synthetic data, similar to data augmentation in images, is commonly believed to improve downstream classification performance. However, current tabular generative methods that learn either the joint distribution $ p(\mathbf{x}, y) $ or the class-conditional distribution $ p(\mathbf{x} \mid y) $ often overfit on small datasets, resulting in poor-quality synthetic data, usually worsening classification performance compared to using real data alone. To solve these challenges, we introduce TabEBM, a novel class-conditional generative method using Energy-Based Models (EBMs). Unlike existing methods that use a shared model to approximate all class-conditional densities, our key innovation is to create distinct EBM generative models for each class, each modelling its class-specific data distribution individually. This approach creates robust energy landscapes, even in ambiguous class distributions. Our experiments show that TabEBM generates synthetic data with higher quality and better statistical fidelity than existing methods. When used for data augmentation, our synthetic data consistently improves the classification performance across diverse datasets of various sizes, especially small ones. Code is available at https://github.com/andreimargeloiu/TabEBM., Comment: Accepted by the Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS 2024)

Coronal mass ejections (CMEs) are complex magnetized plasma structures in which the magnetic field spirals around a central axis, forming what is known as a flux rope (FR). The central FR axis can be oriented at any angle to the ecliptic. Throughout its journey, a CME will encounter interplanetary magnetic field and solar wind which are neither homogeneous nor isotropic. Consequently, CMEs with different orientations will encounter different ambient medium conditions and, thus, the interaction of a CME with its surrounding environment will vary depending on the orientation of its FR axis, among other factors. This study aims to understand the effect of inclination on CME propagation. We performed simulations with the EUHFORIA 3D magnetohydrodynamic model. This study focuses on two CMEs modelled as spheromaks with nearly identical properties, differing only by their inclination. We show the effects of CME orientation on sheath evolution, MHD drag, and non-radial flows by analyzing the model data from a swarm of 81 virtual spacecraft scattered across the inner heliospheric. We have found that the sheath duration increases with radial distance from the Sun and that the rate of increase is greater on the flanks of the CME. Non-radial flows within the studied sheath region appear larger outside the ecliptic plane, indicating a "sliding" of the IMF in the out-of ecliptic plane. We found that the calculated drag parameter does not remain constant with radial distance and that the inclination dependence of the drag parameter can not be resolved with our numerical setup.

We prove that $K_n+I$, the complete graph of an even order with a $1$-factor duplicated, admits a decomposition into $2$-factors, each a disjoint union of cycles of length $m \geq 5$ if and only if $m \mid n$, except possibly when $m$ is odd and $n=4m$. In addition, we show that $K_n+I$ admits a decomposition into $2$-factors, each a disjoint union of cycles of lengths $m_1, \ldots, m_t$, whenever $m_1, \ldots, m_t$ are all even.

The Honeymoon Oberwolfach Problem HOP$(2m_1,2m_2,\ldots,2m_t)$ asks the following question. Given $n=m_1+m_2+\ldots +m_t$ newlywed couples at a conference and $t$ round tables of sizes $2m_1,2m_2,\ldots,2m_t$, is it possible to arrange the $2n$ participants at these tables for $2n-2$ meals so that each participant sits next to their spouse at every meal, and sits next to every other participant exactly once? A solution to HOP$(2m_1,2m_2,\ldots,2m_t)$ is a decomposition of $K_{2n}+(2n-3)I$, the complete graph $K_{2n}$ with $2n-3$ additional copies of a fixed 1-factor $I$, into 2-factors, each consisting of disjoint $I$-alternating cycles of lengths $2m_1,2m_2,\ldots,2m_t$. The Honeymoon Oberwolfach Problem was introduced in a 2019 paper by Lepine and \v{S}ajna. The authors conjectured that HOP$(2m_1,2m_2,\ldots,$ $2m_t)$ has a solution whenever the obvious necessary conditions are satisfied, and proved the conjecture for several large cases, including the uniform cycle length case $m_1=\ldots=m_t$, and the small cases with $n \le 9$. In the present paper, we extend the latter result to all cases with $n \le 20$ using a computer search.

Text embeddings are essential for many tasks, such as document retrieval, clustering, and semantic similarity assessment. In this paper, we study how to contrastively train text embedding models in a compute-optimal fashion, given a suite of pre-trained decoder-only language models. Our innovation is an algorithm that produces optimal configurations of model sizes, data quantities, and fine-tuning methods for text-embedding models at different computational budget levels. The resulting recipe, which we obtain through extensive experiments, can be used by practitioners to make informed design choices for their embedding models. Specifically, our findings suggest that full fine-tuning and low-rank adaptation fine-tuning produce optimal models at lower and higher computational budgets respectively.

Tabular data is prevalent in many critical domains, yet it is often challenging to acquire in large quantities. This scarcity usually results in poor performance of machine learning models on such data. Data augmentation, a common strategy for performance improvement in vision and language tasks, typically underperforms for tabular data due to the lack of explicit symmetries in the input space. To overcome this challenge, we introduce TabMDA, a novel method for manifold data augmentation on tabular data. This method utilises a pre-trained in-context model, such as TabPFN, to map the data into an embedding space. TabMDA performs label-invariant transformations by encoding the data multiple times with varied contexts. This process explores the learned embedding space of the underlying in-context models, thereby enlarging the training dataset. TabMDA is a training-free method, making it applicable to any classifier. We evaluate TabMDA on five standard classifiers and observe significant performance improvements across various tabular datasets. Our results demonstrate that TabMDA provides an effective way to leverage information from pre-trained in-context models to enhance the performance of downstream classifiers. Code is available at https://github.com/AdrianBZG/TabMDA., Comment: Presented at 1st ICML Workshop on In-Context Learning (ICL @ ICML 2024)

Learning holistic computational representations in physical, chemical or biological systems requires the ability to process information from different distributions and modalities within the same model. Thus, the demand for multimodal machine learning models has sharply risen for modalities that go beyond vision and language, such as sequences, graphs, time series, or tabular data. While there are many available multimodal fusion and alignment approaches, most of them require end-to-end training, scale quadratically with the number of modalities, cannot handle cases of high modality imbalance in the training set, or are highly topology-specific, making them too restrictive for many biomedical learning tasks. This paper presents Multimodal Lego (MM-Lego), a modular and general-purpose fusion and model merging framework to turn any set of encoders into a competitive multimodal model with no or minimal fine-tuning. We achieve this by introducing a wrapper for unimodal encoders that enforces lightweight dimensionality assumptions between modalities and harmonises their representations by learning features in the frequency domain to enable model merging with little signal interference. We show that MM-Lego 1) can be used as a model merging method which achieves competitive performance with end-to-end fusion models without any fine-tuning, 2) can operate on any unimodal encoder, and 3) is a model fusion method that, with minimal fine-tuning, achieves state-of-the-art results on six benchmarked multimodal biomedical tasks.

Concept-based explainability methods provide insight into deep learning systems by constructing explanations using human-understandable concepts. While the literature on human reasoning demonstrates that we exploit relationships between concepts when solving tasks, it is unclear whether concept-based methods incorporate the rich structure of inter-concept relationships. We analyse the concept representations learnt by concept-based models to understand whether these models correctly capture inter-concept relationships. First, we empirically demonstrate that state-of-the-art concept-based models produce representations that lack stability and robustness, and such methods fail to capture inter-concept relationships. Then, we develop a novel algorithm which leverages inter-concept relationships to improve concept intervention accuracy, demonstrating how correctly capturing inter-concept relationships can improve downstream tasks., Comment: Accepted at ICML 2024

We compile the catalog of Hvar Observatory solar observations in the time period corresponding to regular digitally stored chromospheric and photospheric observations 2010-2019. We make basic characterisation of observed phenomena and compare them to catalogs which are based on full disc solar images. We compile a catalog of observed ARs consisting of 1100 entries, where each AR is classified according to McIntosh and Mt Wilson classifications. We find that HVAR observations are biased towards more frequently observing more complex ARs and observing them in longer time periods, likely related to the small FOV not encompassing the whole solar disc. In H$\alpha$ observations we catalog conspicuous filaments/prominences and flares. We characterise filaments according to their location, chirality (if possible) and eruptive signatures. Analysis of the eruptive filaments reveals a slight bias in HVAR catalog towards observation of partial eruptions, possibly related to the observers tendency to observe filament which already showed some activity. In the flare catalog we focus on their observed eruptive signatures (loops or ribbons) and their shape. In addition, we associate them to GOES soft X-ray flares to determine their corresponding class. We find that HVAR observations seem biased towards more frequently observing stronger flares and observing them in longer time periods. We demonstrate the feasibility of the catalog on a case study of the flare detected on 2 August 2011 in HVAR H$\alpha$ observations and related Sun-to-Earth phenomena. Through flare-CME-ICME association we demonstrate the agreement of remote and in situ properties. The data used for this study, as well as the catalog, are made publicly available., Comment: Solar Physics, 24 pages, 14 figures

The success of Large Language Models (LLMs), e.g., ChatGPT, is witnessed by their planetary popularity, their capability of human-like communication, and also by their steadily improved reasoning performance. However, it remains unclear whether LLMs reason. It is an open problem how traditional neural networks can be qualitatively extended to go beyond the statistic paradigm and achieve high-level cognition. Here, we present a novel qualitative extension by generalising computational building blocks from vectors to spheres. We propose Sphere Neural Networks (SphNNs) for human-like reasoning through model construction and inspection, and develop SphNN for syllogistic reasoning, a microcosm of human rationality. SphNN is a hierarchical neuro-symbolic Kolmogorov-Arnold geometric GNN, and uses a neuro-symbolic transition map of neighbourhood spatial relations to transform the current sphere configuration towards the target. SphNN is the first neural model that can determine the validity of long-chained syllogistic reasoning in one epoch without training data, with the worst computational complexity of O(N). SphNN can evolve into various types of reasoning, such as spatio-temporal reasoning, logical reasoning with negation and disjunction, event reasoning, neuro-symbolic unification, and humour understanding (the highest level of cognition). All these suggest a new kind of Herbert A. Simon's scissors with two neural blades. SphNNs will tremendously enhance interdisciplinary collaborations to develop the two neural blades and realise deterministic neural reasoning and human-bounded rationality and elevate LLMs to reliable psychological AI. This work suggests that the non-zero radii of spheres are the missing components that prevent traditional deep-learning systems from reaching the realm of rational reasoning and cause LLMs to be trapped in the swamp of hallucination.

We show that the existence of clouds of ultralight particles surrounding black holes during their cosmological history as members of a binary system can leave a measurable imprint on the distribution of masses and orbital eccentricities observable with future gravitational-wave detectors. Notably, we find that for nonprecessing binaries with chirp masses ${\cal M} \lesssim 10\,M_\odot$, formed exclusively in isolation, larger-than-expected values of the eccentricity, i.e. $e\gtrsim 10^{-2}$ at gravitational-wave frequencies $f_{\rm GW} \simeq 10^{-2}\,$Hz, would provide tantalizing evidence for a new particle of mass between $[0.5,2.5] \times 10^{-12}\,$eV in nature. The predicted evolution of the eccentricity can also drastically affect the in-band phase evolution and peak frequency. These results constitute unique signatures of boson clouds of ultralight particles in the dynamics of binary black holes, which will be readily accessible with the Laser Interferometer Space Antenna, as well as future mid-band and Deci-hertz detectors., Comment: v2: 5 pages + 11 pages appendices and references, 11 figures; few typos corrected; added two sections to the appendices; published in PRL

We present MIPS, a novel method for program synthesis based on automated mechanistic interpretability of neural networks trained to perform the desired task, auto-distilling the learned algorithm into Python code. We test MIPS on a benchmark of 62 algorithmic tasks that can be learned by an RNN and find it highly complementary to GPT-4: MIPS solves 32 of them, including 13 that are not solved by GPT-4 (which also solves 30). MIPS uses an integer autoencoder to convert the RNN into a finite state machine, then applies Boolean or integer symbolic regression to capture the learned algorithm. As opposed to large language models, this program synthesis technique makes no use of (and is therefore not limited by) human training data such as algorithms and code from GitHub. We discuss opportunities and challenges for scaling up this approach to make machine-learned models more interpretable and trustworthy., Comment: 24 pages

Concept-based methods explain model predictions using human-understandable concepts. These models require accurate concept predictors, yet the faithfulness of existing concept predictors to their underlying concepts is unclear. In this paper, we investigate the faithfulness of Concept Bottleneck Models (CBMs), a popular family of concept-based architectures, by looking at whether they respect "localities" in datasets. Localities involve using only relevant features when predicting a concept's value. When localities are not considered, concepts may be predicted based on spuriously correlated features, degrading performance and robustness. This work examines how CBM predictions change when perturbing model inputs, and reveals that CBMs may not capture localities, even when independent concepts are localised to non-overlapping feature subsets. Our empirical and theoretical results demonstrate that datasets with correlated concepts may lead to accurate but uninterpretable models that fail to learn localities. Overall, we find that CBM interpretability is fragile, as CBMs occasionally rely upon spurious features, necessitating further research into the robustness of concept predictors., Comment: Previous Version Accepted at NeurIPs 23 XAI in Action Workshop

Technological advances in medical data collection, such as high-throughput genomic sequencing and digital high-resolution histopathology, have contributed to the rising requirement for multimodal biomedical modelling, specifically for image, tabular and graph data. Most multimodal deep learning approaches use modality-specific architectures that are often trained separately and cannot capture the crucial cross-modal information that motivates the integration of different data sources. This paper presents the Hybrid Early-fusion Attention Learning Network (HEALNet): a flexible multimodal fusion architecture, which a) preserves modality-specific structural information, b) captures the cross-modal interactions and structural information in a shared latent space, c) can effectively handle missing modalities during training and inference, and d) enables intuitive model inspection by learning on the raw data input instead of opaque embeddings. We conduct multimodal survival analysis on Whole Slide Images and Multi-omic data on four cancer datasets from The Cancer Genome Atlas (TCGA). HEALNet achieves state-of-the-art performance compared to other end-to-end trained fusion models, substantially improving over unimodal and multimodal baselines whilst being robust in scenarios with missing modalities., Comment: NeurIPS 2024

Autoformalization is the task of translating natural language materials into machine-verifiable formalisations. Progress in autoformalization research is hindered by the lack of a sizeable dataset consisting of informal-formal pairs expressing the same essence. Existing methods tend to circumvent this challenge by manually curating small corpora or using few-shot learning with large language models. But these methods suffer from data scarcity and formal language acquisition difficulty. In this work, we create $\texttt{MMA}$, a large, flexible, multilingual, and multi-domain dataset of informal-formal pairs, by using a language model to translate in the reverse direction, that is, from formal mathematical statements into corresponding informal ones. Experiments show that language models fine-tuned on $\texttt{MMA}$ produce $16-18\%$ of statements acceptable with minimal corrections on the $\texttt{miniF2F}$ and $\texttt{ProofNet}$ benchmarks, up from $0\%$ with the base model. We demonstrate that fine-tuning on multilingual formal data results in more capable autoformalization models even when deployed on monolingual tasks.

We present a theory describing interaction of structured light, such as light carrying orbital angular momentum, with molecules. The light-matter interaction Hamiltonian we derive is expressed through couplings between spherical gradients of the electric field and the (transition) multipole moments of a particle of any non-trivial rotation point group. Our model can therefore accommodate for an arbitrary complexity of the molecular and electric field structure, and can be straightforwardly extended to atoms or nanostructures. Applying this framework to ro-vibrational spectroscopy of molecules, we uncover the general mechanism of angular momentum exchange between the spin and orbital angular momenta of light, molecular rotation and its center-of-mass motion. We show that the non-zero vorticity of Laguerre-Gaussian beams can strongly enhance certain ro-vibrational transitions that are considered forbidden in the case of non-helical light. We discuss the experimental requirements for the observation of these forbidden transitions in state-of-the-art spatially-resolved spectroscopy measurements., Comment: 19 pages, 4 figures

Concept Bottleneck Models (CBMs) tackle the opacity of neural architectures by constructing and explaining their predictions using a set of high-level concepts. A special property of these models is that they permit concept interventions, wherein users can correct mispredicted concepts and thus improve the model's performance. Recent work, however, has shown that intervention efficacy can be highly dependent on the order in which concepts are intervened on and on the model's architecture and training hyperparameters. We argue that this is rooted in a CBM's lack of train-time incentives for the model to be appropriately receptive to concept interventions. To address this, we propose Intervention-aware Concept Embedding models (IntCEMs), a novel CBM-based architecture and training paradigm that improves a model's receptiveness to test-time interventions. Our model learns a concept intervention policy in an end-to-end fashion from where it can sample meaningful intervention trajectories at train-time. This conditions IntCEMs to effectively select and receive concept interventions when deployed at test-time. Our experiments show that IntCEMs significantly outperform state-of-the-art concept-interpretable models when provided with test-time concept interventions, demonstrating the effectiveness of our approach., Comment: Accepted as a spotlight at the Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023)

The directed Oberwolfach problem OP$^\ast(m_1,\ldots,m_k)$ asks whether the complete symmetric digraph $K_n^\ast$, assuming $n=m_1+\ldots +m_k$, admits a decomposition into spanning subdigraphs, each a disjoint union of $k$ directed cycles of lengths $m_1,\ldots,m_k$. We hereby describe a method for constructing a solution to OP$^\ast(m_1,\ldots,m_k)$ given a solution to OP$^\ast(m_1,\ldots,m_\ell)$, for some $\ell

This paper explores a relationship between invariants of certain group actions and the time-reversibility of two-dimensional polynomial differential systems exhibiting a $1:-1$ resonant singularity at the origin. We focus on the connection of time-reversibility with the Sibirsky subvariety of the center (integrability) variety, which encompasses systems possessing a local analytic first integral near the origin. An algorithm for generating the Sibirsky ideal for these systems is proposed and the algebraic properties of the ideal are examined. Furthermore, using a generalization of the concept of time-reversibility we study $n$-dimensional systems with a $1:\zeta:\zeta^2:\dots:\zeta^{n-1}$ resonant singularity at the origin, where $n$ is prime and $\zeta$ is a primitive $n$-th root of unity. We study the invariants of a Lie group action on the parameter space of the system, leveraging the theory of binomial ideals as a fundamental tool for the analysis. Our study reveals intriguing connections between generalized reversibility, invariants, and binomial ideals, shedding light on their complex interrelations.

Space weather is a multidisciplinary research area connecting scientists from across heliophysics domains seeking a coherent understanding of our space environment that can also serve modern life and society's needs. COSPAR's ISWAT (International Space Weather Action Teams) 'clusters' focus attention on different areas of space weather study while ensuring the coupled system is broadly addressed via regular communications and interactions. The ISWAT cluster "H3: Radiation Environment in the Heliosphere" (https://www.iswat-cospar.org/h3) has been working to provide a scientific platform to understand, characterize and predict the energetic particle radiation in the heliosphere with the practical goal of mitigating radiation risks associated with areospace activities, satellite industry and human space explorations. In particular, present approaches help us understand the physical phenomena at large, optimizing the output of multi-viewpoint observations and pushing current models to their limits. In this paper, we review the scientific aspects of the radiation environment in the heliosphere covering four different radiation types: Solar Energetic Particles (SEPs), Ground Level Enhancement (GLE, a type of SEP events with energies high enough to trigger the enhancement of ground-level detectors), Galactic Cosmic Rays (GCRs) and Anomalous Cosmic Rays (ACRs). We focus on related advances in the research community in the past 10-20 years and what we still lack in terms of understanding and predictive capabilities. Finally we also consider some recommendations related to the improvement of both observational and modeling capabilities in the field of space radiation environment.

We investigate the relation between the spatiotemporal evolution of the dimming region and the dominant direction of the filament eruption and CME propagation for the 28 October 2021 X1.0 flare/CME event observed from multiple viewpoints by Solar Orbiter, STEREO-A, SDO, and SOHO. We propose a method to estimate the dominant dimming direction by tracking its area evolution and emphasize its accurate estimation by calculating the surface area of a sphere for each pixel. To determine the early flux rope propagation direction, we perform 3D reconstruction of the CME via graduated cylindrical shell modeling (GCS) and tie-pointing of the filament. The dimming initially expands radially and later shifts southeast. The orthogonal projections of the reconstructed height evolution of the erupting filament onto the solar surface are located in the sector of the dominant dimming growth, while the orthogonal projections of the inner part of GCS reconstruction align with the total dimming area. The filament reaches a maximum speed of $\approx$250 km/s at a height of about $\approx$180 Mm. The direction of its motion is strongly inclined from the radial (64$^\circ$ to the East, 32$^\circ$ to the South). The 50$^\circ$ difference in the 3D direction between the CME and the filament leg closely corresponds to the CME half-width determined from reconstruction, suggesting a potential relation of the reconstructed filament to the associated leg of the CME body. Our findings highlight that the dominant propagation of the dimming growth reflects the direction of the erupting magnetic structure (filament) low in the solar atmosphere, though the filament evolution is not related directly to the direction of the global CME expansion. The overall dimming morphology closely resembles the inner part of the CME reconstruction, validating the use of dimming observations to obtain insight into the CME direction., Comment: 15 pages, 18 figures, accepted to Astronomy & Astrophysics

Background and objective: The usage of machine learning in medical diagnosis and treatment has witnessed significant growth in recent years through the development of computer-aided diagnosis systems that are often relying on annotated medical radiology images. However, the availability of large annotated image datasets remains a major obstacle since the process of annotation is time-consuming and costly. This paper explores how to automatically annotate a database of medical radiology images with regard to their semantic similarity. Material and methods: An automated, unsupervised approach is used to construct a large annotated dataset of medical radiology images originating from Clinical Hospital Centre Rijeka, Croatia, utilising multimodal sources, including images, DICOM metadata, and narrative diagnoses. Several appropriate feature extractors are tested for each of the data sources, and their utility is evaluated using k-means and k-medoids clustering on a representative data subset. Results: The optimal feature extractors are then integrated into a multimodal representation, which is then clustered to create an automated pipeline for labelling a precursor dataset of 1,337,926 medical images into 50 clusters of visually similar images. The quality of the clusters is assessed by examining their homogeneity and mutual information, taking into account the anatomical region and modality representation. Conclusion: The results suggest that fusing the embeddings of all three data sources together works best for the task of unsupervised clustering of large-scale medical data, resulting in the most concise clusters. Hence, this work is the first step towards building a much larger and more fine-grained annotated dataset of medical radiology images.

The container relocation problem is a combinatorial optimisation problem aimed at finding a sequence of container relocations to retrieve all containers in a predetermined order by minimising a given objective. Relocation rules (RRs), which consist of a priority function and relocation scheme, are heuristics commonly used for solving the mentioned problem due to their flexibility and efficiency. Recently, in many real-world problems it is becoming increasingly important to consider energy consumption. However, for this variant no RRs exist and would need to be designed manually. One possibility to circumvent this issue is by applying hyperheuristics to automatically design new RRs. In this study we use genetic programming to obtain priority functions used in RRs whose goal is to minimise energy consumption. We compare the proposed approach with a genetic algorithm from the literature used to design the priority function. The results obtained demonstrate that the RRs designed by genetic programming achieve the best performance.

Variational Autoencoders and their many variants have displayed impressive ability to perform dimensionality reduction, often achieving state-of-the-art performance. Many current methods however, struggle to learn good representations in High Dimensional, Low Sample Size (HDLSS) tasks, which is an inherently challenging setting. We address this challenge by using an ensemble of lightweight VAEs to learn posteriors over subsets of the feature-space, which get aggregated into a joint posterior in a novel divide-and-conquer approach. Specifically, we present an alternative factorisation of the joint posterior that induces a form of implicit data augmentation that yields greater sample efficiency. Through a series of experiments on eight real-world datasets, we show that our method learns better latent representations in HDLSS settings, which leads to higher accuracy in a downstream classification task. Furthermore, we verify that our approach has a positive effect on disentanglement and achieves a lower estimated Total Correlation on learnt representations. Finally, we show that our approach is robust to partial features at inference, exhibiting little performance degradation even with most features missing.

Tabular biomedical data poses challenges in machine learning because it is often high-dimensional and typically low-sample-size (HDLSS). Previous research has attempted to address these challenges via local feature selection, but existing approaches often fail to achieve optimal performance due to their limitation in identifying globally important features and their susceptibility to the co-adaptation problem. In this paper, we propose ProtoGate, a prototype-based neural model for feature selection on HDLSS data. ProtoGate first selects instance-wise features via adaptively balancing global and local feature selection. Furthermore, ProtoGate employs a non-parametric prototype-based prediction mechanism to tackle the co-adaptation problem, ensuring the feature selection results and predictions are consistent with underlying data clusters. We conduct comprehensive experiments to evaluate the performance and interpretability of ProtoGate on synthetic and real-world datasets. The results show that ProtoGate generally outperforms state-of-the-art methods in prediction accuracy by a clear margin while providing high-fidelity feature selection and explainable predictions. Code is available at https://github.com/SilenceX12138/ProtoGate., Comment: Accepted by the Forty-first International Conference on Machine Learning (ICML2024)