arXiv subjects

Elizaveta Semenova

Publications and source records attributed to Elizaveta Semenova.

At least 19 recordsLinked to original sources

arXiv2026

Comparing statistical learning models in wastewater-based epidemiology: An application to norovirus

Wastewater-based epidemiology (WBE) is an increasingly important tool for infectious disease surveillance, but there has been limited direct comparison of modelling approaches for predicting pathogen concentrations across space and time. We compare the predictive performance of six modelling approaches using norovirus in England as a case study, considering 3,232 wastewater samples from 152 sewage treatment works collected between May 2021 and March 2022 as part of the UK Environmental Monitoring for Health Protection programme. The benchmark was a Bayesian spatio-temporal model using Integrated Nested Laplace Approximation (INLA) with the Stochastic Partial Differential Equation approach (SPDE), compared with Lasso regression, Generalised Additive Models (GAM), Bayesian GAM, Extreme Gradient Boosting (XGBoost), and Random Forest. Models were evaluated using 10-fold spatial-block cross-validation with metrics including mean squared error, bias, correlation, empirical coverage of 95% prediction intervals, interval score, and computational cost. Random Forest was the best overall performing model, achieving the best interval score and nearest to nominal empirical coverage (95.3%), while maintaining point prediction accuracy comparable to XGBoost. The INLA-SPDE model also performed well, with near-nominal empirical coverage, the third best interval score, and consistently low bias across all evaluated metrics. Comparing predicted spatio-temporal trends, both models yielded similar spatial patterns but notable differences in uncertainty estimation. Our findings highlight a trade-off between predictive accuracy, uncertainty quantification, and computational efficiency. Ensemble machine learning methods are well suited to rapid prediction, whereas Bayesian geostatistical models remain valuable when probabilistic decision support is a priority for public health surveillance

stat.AP↗

arXiv2026

Lasing from a Quantum-Dot-Like Buried Heterostructure in an InP Nanobeam Cavity

We report lasing from a lithographically defined buried heterostructure with an estimated lateral footprint of (107 nm)^2, embedded in an InP photonic-crystal nanobeam cavity. This represents the smallest laterally confined buried heterostructure gain region from which lasing has been observed. Despite etching of the active region during cavity definition and the associated risk of surface-related nonradiative recombination, optically pumped devices exhibit a clear lasing threshold and a narrow linewidth. By systematically varying the BH size, we investigate how the lasing threshold depends on the active volume under optical pumping. The estimated intrinsic threshold under ideal carrier injection is 57 nW, comparable to values reported for single quantum-dot nanolasers, highlighting the potential of quantum-dot-scale buried heterostructures as deterministic, scalable gain media for nanophotonic lasers.

physics.optics↗

arXiv2026

When Context Compensates for Sparse Event History: AlphaEarth for Spatio-Temporal Point-Process Forecasting

Spatio-temporal point-process models must often generalise across space when local event histories are sparse. We study whether exogenous spatial context can compensate in such regimes. Using a fixed log-Gaussian Cox process backbone, we compare an event-only model with the same model augmented by AlphaEarth embeddings as linear spatial context. We evaluate spatial transfer on emergency medical services (EMS) forecasting across eight held-out regions, fixed forecast anchors, and a sweep over history length $w$, using only AlphaEarth (AE) embeddings available strictly before each anchor. AE improves out-of-region predictive performance across all history regimes, with the largest gains under scarce histories: approximately $2$--$6\times$ multiplicative improvements at $1-2$ weeks, tapering to roughly $10$--$20\%$ at $w=20$--$104$ weeks. These results show that contextual information can substantially stabilise spatially transferred point-process forecasts when event history is limited.

cs.LG↗

arXiv2026

Telecom-band site-controlled quantum dots with engineered low fine-structure splitting

Deterministic quantum light sources emitting at telecom wavelengths with vanishing fine-structure splitting (FSS) are essential components for scalable quantum communication. While self-assembled Stranski-Krastanov (SK) quantum dots (QDs) are high-quality emitters, their random positioning and shape-induced anisotropy typically limit their use in entangled-photon applications. In this work, we demonstrate site-controlled SK growth where InAs/InP QDs nucleate at the symmetric apexes of truncated InP nanopyramids. Confining adatom diffusion to a small, symmetric nucleation area suppresses anisotropic growth, promoting the nucleation of highly symmetric QDs with FSS reduced to values below our statistically validated resolution limit of $9.2~μ$eV. At the same time, lithographically defined nucleation sites enable deterministic control of the QD position, overcoming the limitations of conventional SK growth. The high structural quality of single symmetric QDs is evidenced by the single-photon character of the emission ($g^{(2)}(0)=0.07^{+0.27}_{-0.07}$) spanning the S, C, and L telecom bands, with no evidence of lithography-induced defects affecting emission dynamics. These results demonstrate that tailoring QD symmetry through nanopyramid growth engineering provides a route toward site-controlled emitters suitable for entangled photon generation and integrated quantum photonics devices.

cond-mat.mes-hall↗

arXiv2026

Scalable Bayesian Additive Models for Stellar Flare Detection via Amortized Gaussian Process Inference and Hidden Markov Models

Gaussian Processes (GPs) are a powerful tool for Bayesian time-series modeling, yet their cubic computational cost remains a severe barrier for application to long, high-cadence datasets in astronomy. While specialized scalable solvers like Celerite elegantly reduce this scaling to linear time, repeatedly evaluating the exact likelihood during iterative Bayesian sampling is a bottleneck for developing more complex models, like hierarchical or additive models in which Celerite is only one component. To make this inference computationally tractable, we introduce a generative surrogate framework. By utilizing a Variational Autoencoder (VAE) to learn a compressed representation of the Celerite prior, we map highly correlated stochastic dependencies into a low-dimensional, isotropic manifold. This transition completely bypasses exact covariance operations, shifting the computational burden to a rapid neural network forward pass. Through an extensive simulation study, we show that the generative surrogate accurately reproduces the structural fidelity of exact physical kernels like Celerite. Finally, we demonstrate embedding our VAE approximation into an additive model that combines Celerite and a hidden Markov model (HMM) for stellar flare detection in time series data of stars. We evaluate the joint VAE+HMM architecture against the exact Celerite+HMM framework on empirical astrophysical time series and demonstrate that the proposed methodology achieves significant reductions in computational time, enabling the rigorous, large-scale characterization of stellar flares across massive data archives.

stat.ML↗

arXiv2026

Quantizing Time-Series Models As Dynamical Systems: Trajectory-Based Quantization Sensitivity Score

We introduce the Trajectory-based Quantization Sensitivity Score (TQS), a metric that reframes post-training quantization (PTQ) through the lens of dynamical-systems stability. By modeling the network's rollout as a discrete-time dynamical system, TQS characterizes how quantization-induced errors propagate and amplify over the rollout horizon. Unlike conventional PTQ methods, where sensitivity analysis is often coupled to the quantization procedure, TQS enables a priori sensitivity estimation decoupled from quantizer selection and bit-width assignment. This separation allows for quantization budget planning even for black-box or compiled networks with fused operators. Building on this, we present TQS-PTQ, a flexible mixed-precision framework that requires no calibration data or costly second-order approximations. Our experiments show that a dynamical-systems perspective provides a robust, high-performing pathway for low-precision deployment in resource-constrained settings.

cs.LG↗

arXiv2026

Evaluating AI-based Scientific Knowledge Synthesis with Epidemiological Systematic Reviews

Systematic literature reviews (SLRs) are a demanding and high-stakes form of scientific knowledge synthesis that remains underspecified as an evaluation setting for large language models (LLMs). We introduce AgentSLR, a large-scale evaluation harness comprising an SLR automation workflow and an expert annotated dataset covering 16,248 articles, designed to test LLM capabilities across the stages of SLRs in epidemiology. Reference annotations were derived from peer-reviewed studies on WHO priority pathogens and produced by domain experts. The harness evaluates each review stage as a separate unit with dedicated metrics enabling targeted failure analysis. We evaluated five frontier reasoning models and found that no single model dominated across all tasks, showing sub-task specialisation often hidden by aggregate benchmarks. Structured data extraction is a major bottleneck, with no model exceeding an average field-level F1 of 0.67. Estimated costs vary substantially, by up to 96 times across evaluated models. Documented failure modes suggest that the evaluated models are not yet reliable enough for unsupervised deployment in epidemiology, where findings can inform public policy.

cs.IR↗

arXiv2026

Local droplet etching-assisted quantum dot epitaxy for telecom C-band quantum light emitters

Significant progress in quantum light sources for quantum communication applications requires reproducible and symmetric quantum emitters acting as single-photon sources capable of generating entangled photons on demand at specific telecom wavelengths. Here, we propose telecom-emitting epitaxial quantum dots (QDs) fabricated using the local droplet etching (LDE) approach. The resulting well-defined, low-density ($10^9$/cm$^2$) QDs based on In$_{x}$Ga$_{1-x}$As are formed in symmetric LDE nanoholes (in-plane aspect ratio of 1.14) in In$_{0.52}$Al$_{0.48}$As. Detailed transmission electron microscopy provides comprehensive insight into the structural integrity, interface quality, and compositional profiles of the QDs, which underpin their promising optical properties. Photoluminescence spectroscopy reveals narrow emission lines (0.2 meV) and high optical quality, while second-order autocorrelation measurements confirm clear single-photon emission, with $g^{(2)}(0)=0.07\pm0.02$ under above-band continuous-wave excitation and $g^{(2)}(0)=0.16 \pm 0.18$ under pulsed excitation. Precise numerical modeling, combining multiband $\boldsymbol{k} \cdot \boldsymbol{p}$ and configuration-interaction methods, supports the optical characterization and identifies thermal excitation pathways that explain the persistence of emission up to liquid-nitrogen temperatures. These results highlight the versatility of the LDE approach for integrating new material systems and pave the way toward scalable fabrication of quantum light sources with tailored emission properties.

physics.optics↗

arXiv2026

From Mice to Trains: Amortized Bayesian Inference on Graph Data

Graphs arise across diverse domains, from biology and chemistry to social and information networks, as well as in transportation and logistics. Inference on graph-structured data requires methods that are permutation-invariant, scalable across varying sizes and sparsities, and capable of capturing complex long-range dependencies, making posterior estimation on graph parameters particularly challenging. Amortized Bayesian Inference (ABI) is a simulation-based framework that employs generative neural networks to enable fast, likelihood-free posterior inference. We adapt ABI to graph data to address these challenges to perform inference on node-, edge-, and graph-level parameters. Our approach couples permutation-invariant graph encoders with flexible neural posterior estimators in a two-module pipeline: a summary network maps attributed graphs to fixed-length representations, and an inference network approximates the posterior over parameters. In this setting, several neural architectures can serve as the summary network. In this work we evaluate multiple architectures and assess their performance on controlled synthetic settings and two real-world domains - biology and logistics - in terms of recovery and calibration.

stat.ML↗

arXiv2026

Scalable Spatiotemporal Inference with Biased Scan Attention Transformer Neural Processes

Neural Processes (NPs) are a rapidly evolving class of models designed to directly model the posterior predictive distribution of stochastic processes. While early architectures were developed primarily as a scalable alternative to Gaussian Processes (GPs), modern NPs tackle far more complex and data-hungry applications spanning geology, epidemiology, climate, and robotics. These applications have placed increasing pressure on the scalability of these models, with many architectures compromising accuracy for scalability. In this paper, we demonstrate that this trade-off is often unnecessary, particularly when modeling fully or partially translation-invariant processes. We propose a versatile new architecture, the Biased Scan Attention Transformer Neural Process (BSA-TNP), which introduces Kernel Regression Blocks (KRBlocks), group-invariant attention biases, and memory-efficient Biased Scan Attention (BSA). BSA-TNP is able to: (1) match or exceed the accuracy of the best models while often training in a fraction of the time, (2) exhibit translation invariance, enabling learning at multiple resolutions simultaneously, (3) transparently model processes that evolve in both space and time, (4) support high-dimensional fixed effects, and (5) scale gracefully, running inference on over 1M test points and 100K context points in under a minute on a single 24GB GPU. Code is provided as part of the `dl4bi` package.

cs.LG↗

arXiv2026

Tokenised Flow Matching for Hierarchical Simulation Based Inference

The cost of simulator evaluations is a key practical bottleneck for Simulation Based Inference (SBI). In hierarchical settings with shared global parameters and exchangeable site-level parameters and observations, this structure can be exploited to improve simulation efficiency. Existing hierarchical SBI approaches factorise the posterior yet still simulate across multiple sites per training sample; We instead explore likelihood factorisation (LF) to train from single-site simulations. In LF sampling we learn a per-site neural surrogate of the simulator and then assemble synthetic multi-site observations to amortise inference for the full hierarchical posterior. Building on this, we propose Tokenised Flow Matching for Posterior Estimation (TFMPE), a tokenised flow matching approach that supports function-valued observations through likelihood factorisation. To enable systematic evaluation, we introduce a benchmark for hierarchical SBI. We validate TFMPE on this benchmark and on realistic infectious disease and computational fluid dynamics models, finding well-calibrated posteriors while reducing computational cost.

cs.LG↗

arXiv2026

Transformer Neural Processes - Kernel Regression

Neural Processes (NPs) are a rapidly evolving class of models designed to directly model the posterior predictive distribution of stochastic processes. Originally developed as a scalable alternative to Gaussian Processes (GPs), which are limited by $O(n^3)$ runtime complexity, the most accurate modern NPs can often rival GPs but still suffer from an $O(n^2)$ bottleneck due to their attention mechanism. We introduce the Transformer Neural Process - Kernel Regression (TNP-KR), a scalable NP featuring: (1) a Kernel Regression Block (KRBlock), a simple, extensible, and parameter efficient transformer block with complexity $O(n_c^2 + n_c n_t)$, where $n_c$ and $n_t$ are the number of context and test points, respectively; (2) a kernel-based attention bias; and (3) two novel attention mechanisms: scan attention (SA), a memory-efficient scan-based attention that when paired with a kernel-based bias can make TNP-KR translation invariant, and deep kernel attention (DKA), a Performer-style attention that implicitly incoporates a distance bias and further reduces complexity to $O(n_c)$. These enhancements enable both TNP-KR variants to perform inference with 100K context points on over 1M test points in under a minute on a single 24GB GPU. On benchmarks spanning meta regression, Bayesian optimization, image completion, and epidemiology, TNP-KR with DKA outperforms its Performer counterpart on nearly every benchmark, while TNP-KR with SA achieves state-of-the-art results.

cs.LG↗

arXiv2026

DeepRV: Accelerating Spatiotemporal Inference with Pre-trained Neural Priors

Gaussian Processes (GPs) provide a flexible and statistically principled foundation for modelling spatiotemporal phenomena, but their $O(N^3)$ scaling makes them intractable for large datasets. Approximate methods such as variational inference (VI), inducing-point (sparse) GPs, low-rank kernel approximations (e.g., Nystrom methods and random Fourier features), and approximations such as INLA improve scalability but typically trade off accuracy, calibration, or modelling flexibility. We introduce DeepRV, a neural-network surrogate that replaces GP prior sampling, while closely matching full GP accuracy at inference including hyperparameter estimates, and reducing computational complexity to $O(N^2)$, increasing scalability and inference speed. DeepRV serves as a drop-in replacement for GP prior realisations in e.g. MCMC-based probabilistic programming pipelines, preserving full model flexibility. Across simulated benchmarks, non-separable spatiotemporal GPs, and a real-world application to education deprivation in London (n = 4,994 locations), DeepRV achieves the highest fidelity to exact GPs while substantially accelerating inference. Code is provided in the dl4bi Python package, with all experiments run on a single consumer-grade GPU to ensure accessibility for practitioners.

stat.ML↗

arXiv2025

SMRS: advocating a unified reporting standard for surrogate models in the artificial intelligence era

Surrogate models are widely used to approximate complex systems across science and engineering to reduce computational costs. Despite their widespread adoption, the field lacks standardisation across key stages of the modelling pipeline, including data sampling, model selection, evaluation, and downstream analysis. This fragmentation limits reproducibility and cross-domain utility -- a challenge further exacerbated by the rapid proliferation of AI-driven surrogate models. We argue for the urgent need to establish a structured reporting standard, the Surrogate Model Reporting Standard (SMRS), that systematically captures essential design and evaluation choices while remaining agnostic to implementation specifics. By promoting a standardised yet flexible framework, we aim to improve the reliability of surrogate modelling, foster interdisciplinary knowledge transfer, and, as a result, accelerate scientific progress in the AI era.

stat.CO↗

arXiv2025

The World Wide recipe: A community-centred framework for fine-grained data collection and regional bias operationalisation

We introduce the World Wide recipe, which sets forth a framework for culturally aware and participatory data collection, and the resultant regionally diverse World Wide Dishes evaluation dataset. We also analyse bias operationalisation to highlight how current systems underperform across several dimensions: (in-)accuracy, (mis-)representation, and cultural (in-)sensitivity, with evidence from qualitative community-based observations and quantitative automated tools. We find that these T2I models generally do not produce quality outputs of dishes specific to various regions. This is true even for the US, which is typically considered more well-resourced in training data -- although the generation of US dishes does outperform that of the investigated African countries. The models demonstrate the propensity to produce inaccurate and culturally misrepresentative, flattening, and insensitive outputs. These representational biases have the potential to further reinforce stereotypes and disproportionately contribute to erasure based on region. The dataset and code are available at https://github.com/oxai/world-wide-dishes.

cs.CY↗

arXiv2025

Real-time small area estimation of food security in Zimbabwe: integrating mobile-phone and face-to-face surveys using joint multilevel regression and poststratification

Real-time, fine-grained monitoring of food security is essential for enabling timely and targeted interventions, thereby supporting the global goal of achieving zero hunger - a key objective of the 2030 Agenda for Sustainable Development. Mobile phone surveys provide a scalable and temporally rich data source that can be tailored to different administrative levels. However, due to cost and operational constraints, maintaining high-frequency data collection while ensuring representativeness at lower administrative levels is often infeasible. We propose a joint multilevel regression and poststratification (jMRP) approach that combines high-frequency and up-to-date mobile phone survey data, designed for higher administrative levels, with an annual face-to-face survey representative at lower levels to produce reliable food security estimates at spatially and temporally finer scales than those originally targeted by the surveys. This methodology accounts for systematic differences in survey responses due to modality and socio-economic characteristics, reducing both sampling and modality bias. We implement the approach in a fully Bayesian manner to quantify uncertainty. We demonstrate the effectiveness of our method using data from Zimbabwe, thus offering a cost-effective solution for real-time monitoring and strengthening decision-making in resource-constrained settings.

stat.AP↗

arXiv2025

Monolithic Integration of Sub-50 nm III-V Nano-Heterostructures on Si (001) for Telecom Photonics

The demand for advanced photonics technology is increasing rapidly, fueled by the necessity for high-performance and cost-effective optical information processing systems extending into the quantum domain. Silicon, benefiting from its mature fabrication processes, stands as an ideal platform. However, its inherent indirect bandgap leads to inefficient light emission. The integration of III-V materials has been proven essential to overcome this drawback. These materials are recognized for their efficient light emission and superior bandgap engineering capabilities, making them indispensable in photonics and beyond. Here, we present the monolithic integration of small-volume III-V nanoheterostructures with silicon via selective area epitaxy in the pyramidal openings etched in (100)-oriented silicon substrate. The precise positioning of the nano-heterostructures is achieved using electron beam lithography. Our atomic resolution imaging and chemical analysis confirms the epitaxial nature of InP growth, revealing well-defined heterointerfaces. Each structure incorporates an InAsP quantum dot-like active medium, and the correlation of the growth parameters with the nanoscale structure was analyzed using advanced electron microscopy. The eight-band k.p calculations demonstrate energy level quantization in three spatial dimensions. Optical characterization shows that heterostructure emission can be engineered to cover the entire telecom wavelength range. Consequently, these InAsP/InP nano-heterostructures could serve as a gain medium for silicon-based hybrid nano-lasers and nano-LEDs and quantum light sources in the telecom wavelength range.

physics.optics↗

arXiv2025

Efficient passivation of III-As(P) photonic interfaces

Surface effects can significantly impact the performance of nanophotonic and quantum photonic devices, especially as the device dimensions are reduced. In this work, we propose and investigate a novel approach to surface passivation to mitigate these challenges in photonic nanostructures with III-As(P) quantum wells defined by a dry etching process. The nanostructures are annealed under the phosphine (PH$_3$) ambient inside a metal-organic vapor phase epitaxy chamber to eliminate surface and subsurface defects induced during the dry etching and subsequent oxidation of the etched sidewalls. Moreover, encapsulation of the active material with a wider bandgap material allows for maintaining the band structure of the device, mitigating band bending effects. Our findings reveal an almost order of magnitude reduction in the surface recombination velocity from $2 \times 10^3 \, \mathrm{cm/s}$ for the PH$_3$ annealing compared to $1.5 \times 10^4 \, \mathrm{cm/s}$ for the non-passivated structures and $5 \times 10^3 \, \mathrm{cm/s}$ for the standard method based on (NH$_4$)$_2$S wet treatment followed by Al$_2$O$_3$ encapsulation. A further reduction to $5 \times 10^2 \, \mathrm{cm/s}$ is achieved for the InP-regrown samples. Additionally, we develop a model accounting for the impact of surface charges in the analysis of time-resolved photoluminescence curves and demonstrate that the proposed passivation method effectively reduces the surface charge density on the sidewalls of the studied quantum well-based photonic nanostructures.

physics.optics↗