CyberSec Research

Standard attention mechanisms in transformers employ static token representations that remain unchanged across all pair-wise computations in each layer. This limits their representational alignment with the potentially diverse relational dynamics of each token-pair interaction. While they excel in domains with relatively homogeneous relationships, standard attention's static relational learning struggles to capture the diverse, heterogeneous inter-channel dependencies of multivariate time series (MTS) data--where different channel-pair interactions within a single system may be governed by entirely different physical laws or temporal dynamics. To better align the attention mechanism for such domain phenomena, we propose attention with dynamic relational priming (prime attention). Unlike standard attention where each token presents an identical representation across all of its pair-wise interactions, prime attention tailors each token dynamically (or per interaction) through learnable modulations to best capture the unique relational dynamics of each token pair, optimizing each pair-wise interaction for that specific relationship. This representational plasticity of prime attention enables effective extraction of relationship-specific information in MTS while maintaining the same asymptotic computational complexity as standard attention. Our results demonstrate that prime attention consistently outperforms standard attention across benchmarks, achieving up to 6.5\% improvement in forecasting accuracy. In addition, we find that prime attention achieves comparable or superior performance using up to 40\% less sequence length compared to standard attention, further demonstrating its superior relational modeling capabilities.

The study of neural representations, both in biological and artificial systems, is increasingly revealing the importance of geometric and topological structures. Inspired by this, we introduce Event2Vec, a novel framework for learning representations of discrete event sequences. Our model leverages a simple, additive recurrent structure to learn composable, interpretable embeddings. We provide a theoretical analysis demonstrating that, under specific training objectives, our model's learned representations in a Euclidean space converge to an ideal additive structure. This ensures that the representation of a sequence is the vector sum of its constituent events, a property we term the linear additive hypothesis. To address the limitations of Euclidean geometry for hierarchical data, we also introduce a variant of our model in hyperbolic space, which is naturally suited to embedding tree-like structures with low distortion. We present experiments to validate our hypothesis and demonstrate the benefits of each geometry, highlighting the improved performance of the hyperbolic model on hierarchical event sequences.

Novel view synthesis (NVS) of in-the-wild garments is a challenging task due significant occlusions, complex human poses, and cloth deformations. Prior methods rely on synthetic 3D training data consisting of mostly unoccluded and static objects, leading to poor generalization on real-world clothing. In this paper, we propose HoloGarment (Hologram-Garment), a method that takes 1-3 images or a continuous video of a person wearing a garment and generates 360{\deg} novel views of the garment in a canonical pose. Our key insight is to bridge the domain gap between real and synthetic data with a novel implicit training paradigm leveraging a combination of large-scale real video data and small-scale synthetic 3D data to optimize a shared garment embedding space. During inference, the shared embedding space further enables dynamic video-to-360{\deg} NVS through the construction of a garment "atlas" representation by finetuning a garment embedding on a specific real-world video. The atlas captures garment-specific geometry and texture across all viewpoints, independent of body pose or motion. Extensive experiments show that HoloGarment achieves state-of-the-art performance on NVS of in-the-wild garments from images and videos. Notably, our method robustly handles challenging real-world artifacts -- such as wrinkling, pose variation, and occlusion -- while maintaining photorealism, view consistency, fine texture details, and accurate geometry. Visit our project page for additional results: https://johannakarras.github.io/HoloGarment

Model selection in non-linear models often prioritizes performance metrics over statistical tests, limiting the ability to account for sampling variability. We propose the use of a statistical test to assess the equality of variances in forecasting errors. The test builds upon the classic Morgan-Pitman approach, incorporating enhancements to ensure robustness against data with heavy-tailed distributions or outliers with high variance, plus a strategy to make residuals from machine learning models statistically independent. Through a series of simulations and real-world data applications, we demonstrate the test's effectiveness and practical utility, offering a reliable tool for model evaluation and selection in diverse contexts.

Generative models for materials, especially inorganic crystals, hold potential to transform the theoretical prediction of novel compounds and structures. Advancement in this field depends critically on robust benchmarks and minimal, information-rich datasets that enable meaningful model evaluation. This paper critically examines common datasets and reported metrics for a crystal structure prediction task$\unicode{x2014}$generating the most likely structures given the chemical composition of a material. We focus on three key issues: First, materials datasets should contain unique crystal structures; for example, we show that the widely-utilized carbon-24 dataset only contains $\approx$40% unique structures. Second, materials datasets should not be split randomly if polymorphs of many different compositions are numerous, which we find to be the case for the perov-5 dataset. Third, benchmarks can mislead if used uncritically, e.g., reporting a match rate metric without considering the structural variety exhibited by identical building blocks. To address these oft-overlooked issues, we introduce several fixes. We provide revised versions of the carbon-24 dataset: one with duplicates removed, one deduplicated and split by number of atoms $N$, and two containing only identical structures but with different unit cells. We also propose a new split for the perov-5 dataset which ensures polymorphs are grouped within each split subset, setting a more sensible standard for benchmarking model performance. Finally, we present METRe and cRMSE, new model evaluation metrics that can correct existing issues with the match rate metric.

Human face synthesis and manipulation are increasingly important in entertainment and AI, with a growing demand for highly realistic, identity-preserving images even when only unpaired, unaligned datasets are available. We study unpaired face manipulation via adversarial learning, moving from autoencoder baselines to a robust, guided CycleGAN framework. While autoencoders capture coarse identity, they often miss fine details. Our approach integrates spectral normalization for stable training, identity- and perceptual-guided losses to preserve subject identity and high-level structure, and landmark-weighted cycle constraints to maintain facial geometry across pose and illumination changes. Experiments show that our adversarial trained CycleGAN improves realism (FID), perceptual quality (LPIPS), and identity preservation (ID-Sim) over autoencoders, with competitive cycle-reconstruction SSIM and practical inference times, which achieved high quality without paired datasets and approaching pix2pix on curated paired subsets. These results demonstrate that guided, spectrally normalized CycleGANs provide a practical path from autoencoders to robust unpaired face manipulation.

Multivariate longitudinal data of mixed-type are increasingly collected in many science domains. However, algorithms to cluster this kind of data remain scarce, due to the challenge to simultaneously model the within- and between-time dependence structures for multivariate data of mixed kind. We introduce the Mixture of Mixed-Matrices (MMM) model: reorganizing the data in a three-way structure and assuming that the non-continuous variables are observations of underlying latent continuous variables, the model relies on a mixture of matrix-variate normal distributions to perform clustering in the latent dimension. The MMM model is thus able to handle continuous, ordinal, binary, nominal and count data and to concurrently model the heterogeneity, the association among the responses and the temporal dependence structure in a parsimonious way and without assuming conditional independence. The inference is carried out through an MCMC-EM algorithm, which is detailed. An evaluation of the model through synthetic data shows its inference abilities. A real-world application on financial data is presented.

The first part of this paper studies the evolution of gradient flow for homogeneous neural networks near a class of saddle points exhibiting a sparsity structure. The choice of these saddle points is motivated from previous works on homogeneous networks, which identified the first saddle point encountered by gradient flow after escaping the origin. It is shown here that, when initialized sufficiently close to such saddle points, gradient flow remains near the saddle point for a sufficiently long time, during which the set of weights with small norm remain small but converge in direction. Furthermore, important empirical observations are made on the behavior of gradient descent after escaping these saddle points. The second part of the paper, motivated by these results, introduces a greedy algorithm to train deep neural networks called Neuron Pursuit (NP). It is an iterative procedure which alternates between expanding the network by adding neuron(s) with carefully chosen weights, and minimizing the training loss using this augmented network. The efficacy of the proposed algorithm is validated using numerical experiments.

We present a learnable physics simulator that provides accurate motion and force-torque prediction of robot end effectors in contact-rich manipulation. The proposed model extends the state-of-the-art GNN-based simulator (FIGNet) with novel node and edge types, enabling action-conditional predictions for control and state estimation tasks. In simulation, the MPC agent using our model matches the performance of the same controller with the ground truth dynamics model in a challenging peg-in-hole task, while in the real-world experiment, our model achieves a 50% improvement in motion prediction accuracy and 3$\times$ increase in force-torque prediction precision over the baseline physics simulator. Source code and data are publicly available.

Climate change is accelerating the frequency and severity of unprecedented events, deviating from established patterns. Predicting these out-of-distribution (OOD) events is critical for assessing risks and guiding climate adaptation. While machine learning (ML) models have shown promise in providing precise, high-speed climate predictions, their ability to generalize under distribution shifts remains a significant limitation that has been underexplored in climate contexts. This research systematically evaluates state-of-the-art ML-based climate models in diverse OOD scenarios by adapting established OOD evaluation methodologies to climate data. Experiments on large-scale datasets reveal notable performance variability across scenarios, shedding light on the strengths and limitations of current models. These findings underscore the importance of robust evaluation frameworks and provide actionable insights to guide the reliable application of ML for climate risk forecasting.

Actor-critic algorithms for deep multi-agent reinforcement learning (MARL) typically employ a policy update that responds to the current strategies of other agents. While being straightforward, this approach does not account for the updates of other agents at the same update step, resulting in miscoordination. In this paper, we introduce the $K$-Level Policy Gradient (KPG), a method that recursively updates each agent against the updated policies of other agents, speeding up the discovery of effective coordinated policies. We theoretically prove that KPG with finite iterates achieves monotonic convergence to a local Nash equilibrium under certain conditions. We provide principled implementations of KPG by applying it to the deep MARL algorithms MAPPO, MADDPG, and FACMAC. Empirically, we demonstrate superior performance over existing deep MARL algorithms in StarCraft II and multi-agent MuJoCo.

Deep learning (DL) methods are widely used to extract high-dimensional patterns from the sequence features of radar echo signals. However, conventional DL algorithms face challenges such as redundant feature segments, and constraints from restricted model sizes. To address these issues, we propose a framework that integrates feature preprocessing with large language models (LLMs). Our preprocessing module tokenizes radar sequence features, applies a patch selection algorithm to filter out uninformative segments, and projects the selected patches into embeddings compatible with the feature space of pre-trained LLMs. Leveraging these refined embeddings, we incorporate a pre-trained LLM, fine-tuning only the normalization layers to reduce training burdens while enhancing performance. Experiments on measured datasets demonstrate that the proposed method significantly outperforms the state-of-the-art baselines on supervised learning tests.

Graph machine learning models often achieve similar overall performance yet behave differently at the node level, failing on different subsets of nodes with varying reliability. Standard evaluation metrics such as accuracy obscure these fine grained differences, making it difficult to diagnose when and where models fail. We introduce NodePro, a node profiling framework that enables fine-grained diagnosis of model behavior by assigning interpretable profile scores to individual nodes. These scores combine data-centric signals, such as feature dissimilarity, label uncertainty, and structural ambiguity, with model-centric measures of prediction confidence and consistency during training. By aligning model behavior with these profiles, NodePro reveals systematic differences between models, even when aggregate metrics are indistinguishable. We show that node profiles generalize to unseen nodes, supporting prediction reliability without ground-truth labels. Finally, we demonstrate the utility of NodePro in identifying semantically inconsistent or corrupted nodes in a structured knowledge graph, illustrating its effectiveness in real-world settings.

This paper proposes deception as a mechanism for out-of-distribution (OOD) generalization: by learning data representations that make training data appear independent and identically distributed (iid) to an observer, we can identify stable features that eliminate spurious correlations and generalize to unseen domains. We refer to this principle as deceptive risk minimization (DRM) and instantiate it with a practical differentiable objective that simultaneously learns features that eliminate distribution shifts from the perspective of a detector based on conformal martingales while minimizing a task-specific loss. In contrast to domain adaptation or prior invariant representation learning methods, DRM does not require access to test data or a partitioning of training data into a finite number of data-generating domains. We demonstrate the efficacy of DRM on numerical experiments with concept shift and a simulated imitation learning setting with covariate shift in environments that a robot is deployed in.

This study introduces Universal Delay Embedding (UDE), a pretrained foundation model designed to revolutionize time-series forecasting through principled integration of delay embedding representation and Koopman operator prediction. Leveraging Takens' embedding theorem, UDE as a dynamical representation of observed data constructs two-dimensional subspace patches from Hankel matrices, theoretically preserving dynamical and topological properties of underlying dynamical systems. Such patches are viewed as images, which can be efficiently processed by exploiting advanced deep learning technologies. Computationally, these patches further serve as tokens for learning a self-attention encoder, thus enabling accurate prediction of nonlinear time-series by a finite-dimensional Koopman operator in a linear manner in a latent space. Extensive evaluations across various benchmarks and real-world climate datasets demonstrate over 20% average reduction in mean squared error versus state-of-the-art foundation models, alongside superior generalization in fine-tuning scenarios. In particular, the learned dynamical representations and Koopman operator prediction forms from the patches exhibit exceptional interpretability, with consistent identification of topologically informative subspaces and robust encoding of domain-invariant dynamics, establishing UDE as a scalable, interpretable framework for universal time-series modeling and forecasting with broad scientific and industrial applicability.

Branched broomrape (Phelipanche ramosa) is a chlorophyll-deficient parasitic weed that threatens tomato production by extracting nutrients from the host. We investigate early detection using leaf-level spectral reflectance (400-2500 nm) and ensemble machine learning. In a field experiment in Woodland, California, we tracked 300 tomato plants across growth stages defined by growing degree days (GDD). Leaf reflectance was acquired with a portable spectrometer and preprocessed (band denoising, 1 nm interpolation, Savitzky-Golay smoothing, correlation-based band reduction). Clear class differences were observed near 1500 nm and 2000 nm water absorption features, consistent with reduced leaf water content in infected plants at early stages. An ensemble combining Random Forest, XGBoost, SVM with RBF kernel, and Naive Bayes achieved 89% accuracy at 585 GDD, with recalls of 0.86 (infected) and 0.93 (noninfected). Accuracy declined at later stages (e.g., 69% at 1568 GDD), likely due to senescence and weed interference. Despite the small number of infected plants and environmental confounders, results show that proximal sensing with ensemble learning enables timely detection of broomrape before canopy symptoms are visible, supporting targeted interventions and reduced yield losses.

Decision trees are a ubiquitous model for classification and regression tasks due to their interpretability and efficiency. However, solving the optimal decision tree (ODT) problem remains a challenging combinatorial optimization task. Even for the simplest splitting rules--axis-parallel hyperplanes--it is NP-hard to optimize. In Part I of this series, we rigorously defined the proper decision tree model through four axioms and, based on these, introduced four formal definitions of the ODT problem. From these definitions, we derived four generic algorithms capable of solving ODT problems for arbitrary decision trees satisfying the axioms. We also analyzed the combinatorial geometric properties of hypersurfaces, showing that decision trees defined by polynomial hypersurface splitting rules satisfy the proper axioms that we proposed. In this second paper (Part II) of this two-part series, building on the algorithmic and geometric foundations established in Part I, we introduce the first hypersurface decision tree (HODT) algorithm. To the best of our knowledge, existing optimal decision tree methods are, to date, limited to hyperplane splitting rules--a special case of hypersurfaces--and rely on general-purpose solvers. In contrast, our HODT algorithm addresses the general hypersurface decision tree model without requiring external solvers. Using synthetic datasets generated from ground-truth hyperplane decision trees, we vary tree size, data size, dimensionality, and label and feature noise. Results showing that our algorithm recovers the ground truth more accurately than axis-parallel trees and exhibits greater robustness to noise. We also analyzed generalization performance across 30 real-world datasets, showing that HODT can achieve up to 30% higher accuracy than the state-of-the-art optimal axis-parallel decision tree algorithm when tree complexity is properly controlled.

Modern tensor applications, especially foundation models and generative AI applications require multiple input modalities (both vision and language), which increases the demand for flexible accelerator architecture. Existing frameworks suffer from the trade-off between design flexibility and productivity of RTL generation: either limited to very few hand-written templates or cannot automatically generate the RTL. To address this challenge, we propose the LEGO framework, which targets tensor applications and automatically generates spatial architecture design and outputs synthesizable RTL code without handwritten RTL design templates. Leveraging the affine-transformation-based architecture representation, LEGO front end finds interconnections between function units, synthesizes the memory system, and fuses different spatial dataflow designs based on data reuse analysis. LEGO back end then translates the hardware in a primitive-level graph to perform lower-level optimizations, and applies a set of linear-programming algorithms to optimally insert pipeline registers and reduce the overhead of unused logic when switching spatial dataflows. Our evaluation demonstrates that LEGO can achieve 3.2x speedup and 2.4x energy efficiency compared to previous work Gemmini, and can generate one architecture for diverse modern foundation models in generative AI applications.