arXiv subjects

Andreas Hellander

Publications and source records attributed to Andreas Hellander.

At least 19 recordsLinked to original sources

arXiv2026

GeoFlowVLM: Geometry-Aware Joint Uncertainty for Frozen Vision-Language Embedding

Standard dual-encoder vision-language models that map images and text to deterministic points on a shared unit hypersphere through $\ell_2$ normalization typically expose neither \emph{aleatoric} uncertainty (cross-modal ambiguity) nor \emph{epistemic} uncertainty (lack of training-distribution support). Existing post-hoc methods either recover at most one of the two uncertainty components, or ignore the hyperspherical geometry of these models' embeddings. We propose \textbf{GeoFlowVLM} as a post-hoc adapter that learns the joint distribution of paired $\ell_2$-normalised dual-encoder VLM embeddings on the product hypersphere $\mathbb{S}^{d-1} \times \mathbb{S}^{d-1}$ via Riemannian flow matching with a single masked velocity field. A consistency result shows that, in the population limit, the trained network exposes the joint flow and both cross-modal conditional flows as valid Riemannian flow-matching velocity fields on their respective domains. We derive two quantities from this single model: a conditional retrieval entropy that quantifies aleatoric ambiguity with a decision-theoretic interpretation via a Fano-type bound, and a marginal-typicality epistemic score justified by an exact chain-rule decomposition of the joint NLL. This decomposition isolates a cross-modal pointwise-mutual-information term that is structurally discriminative rather than epistemic, and is empirically the only consistently uninformative standalone component. Empirically, the entropy tracks Recall@1 with near-ideal monotonic calibration across three retrieval benchmarks in both directions, and the marginal-typicality sum yields consistently calibrated selective accuracy across four zero-shot classification benchmarks.

cs.LG↗

arXiv2026

OneFlowSBI: One Model, Many Queries for Simulation-Based Inference

We introduce \textit{OneFlowSBI}, a unified framework for simulation-based inference that learns a single flow-matching generative model over the joint distribution of parameters and observations. Leveraging a query-aware masking distribution during training, the same model supports multiple inference tasks, including posterior sampling, likelihood estimation, and arbitrary conditional distributions, without task-specific retraining. We evaluate \textit{OneFlowSBI} on ten benchmark inference problems and two high-dimensional real-world inverse problems across multiple simulation budgets. \textit{OneFlowSBI} is shown to deliver competitive performance against state-of-the-art generalized inference solvers and specialized posterior estimators, while enabling efficient sampling with few ODE integration steps and remaining robust under noisy and partially observed data.

stat.ML↗

arXiv2026

Epistemic Uncertainty Quantification for Pre-trained VLMs via Riemannian Flow Matching

Vision-Language Models (VLMs) are typically deterministic in nature and lack intrinsic mechanisms to quantify epistemic uncertainty, which reflects the model's lack of knowledge or ignorance of its own representations. We theoretically motivate negative log-density of an embedding as a proxy for the epistemic uncertainty, where low-density regions signify model ignorance. The proposed method REPVLM computes the probability density on the hyperspherical manifold of the VLM embeddings using Riemannian Flow Matching. We empirically demonstrate that REPVLM achieves near-perfect correlation between uncertainty and prediction error, significantly outperforming existing baselines. Beyond classification, we also demonstrate that the model also provides a scalable metric for out-of-distribution detection and automated data curation.

cs.LG↗

arXiv2025

Practical Feasibility of Gradient Inversion Attacks in Federated Learning

Gradient inversion attacks are often presented as a serious privacy threat in federated learning, with recent work reporting increasingly strong reconstructions under favorable experimental settings. However, it remains unclear whether such attacks are feasible in modern, performance-optimized systems deployed in practice. In this work, we evaluate the practical feasibility of gradient inversion for image-based federated learning. We conduct a systematic study across multiple datasets and tasks, including image classification and object detection, using canonical vision architectures at contemporary resolutions. Our results show that while gradient inversion remains possible for certain legacy or transitional designs under highly restrictive assumptions, modern, performance-optimized models consistently resist meaningful reconstruction visually. We further demonstrate that many reported successes rely on upper-bound settings, such as inference mode operation or architectural simplifications which do not reflect realistic training pipelines. Taken together, our findings indicate that, under an honest-but-curious server assumption, high-fidelity image reconstruction via gradient inversion does not constitute a critical privacy risk in production-optimized federated learning systems, and that practical risk assessments must carefully distinguish diagnostic attack settings from real-world deployments.

cs.CR↗

arXiv2025

Exploiting the Asymmetric Uncertainty Structure of Pre-trained VLMs on the Unit Hypersphere

Vision-language models (VLMs) as foundation models have significantly enhanced performance across a wide range of visual and textual tasks, without requiring large-scale training from scratch for downstream tasks. However, these deterministic VLMs fail to capture the inherent ambiguity and uncertainty in natural language and visual data. Recent probabilistic post-hoc adaptation methods address this by mapping deterministic embeddings onto probability distributions; however, existing approaches do not account for the asymmetric uncertainty structure of the modalities, and the constraint that meaningful deterministic embeddings reside on a unit hypersphere, potentially leading to suboptimal performance. In this paper, we address the asymmetric uncertainty structure inherent in textual and visual data, and propose AsymVLM to build probabilistic embeddings from pre-trained VLMs on the unit hypersphere, enabling uncertainty quantification. We validate the effectiveness of the probabilistic embeddings on established benchmarks, and present comprehensive ablation studies demonstrating the inherent nature of asymmetry in the uncertainty structure of textual and visual data.

cs.LG↗

arXiv2025

ConDiSim: Conditional Diffusion Models for Simulation Based Inference

We present a conditional diffusion model - ConDiSim, for simulation-based inference of complex systems with intractable likelihoods. ConDiSim leverages denoising diffusion probabilistic models to approximate posterior distributions, consisting of a forward process that adds Gaussian noise to parameters, and a reverse process learning to denoise, conditioned on observed data. This approach effectively captures complex dependencies and multi-modalities within posteriors. ConDiSim is evaluated across ten benchmark problems and two real-world test problems, where it demonstrates effective posterior approximation accuracy while maintaining computational efficiency and stability in model training. ConDiSim offers a robust and extensible framework for simulation-based inference, particularly suitable for parameter inference workflows requiring fast inference methods.

cs.LG↗

arXiv2024

Variational Autoencoders for Efficient Simulation-Based Inference

We present a generative modeling approach based on the variational inference framework for likelihood-free simulation-based inference. The method leverages latent variables within variational autoencoders to efficiently estimate complex posterior distributions arising from stochastic simulations. We explore two variations of this approach distinguished by their treatment of the prior distribution. The first model adapts the prior based on observed data using a multivariate prior network, enhancing generalization across various posterior queries. In contrast, the second model utilizes a standard Gaussian prior, offering simplicity while still effectively capturing complex posterior distributions. We demonstrate the ability of the proposed approach to approximate complex posteriors while maintaining computational efficiency on well-established benchmark problems.

cs.LG↗

arXiv2023

Toward efficient resource utilization at edge nodes in federated learning

Federated learning (FL) enables edge nodes to collaboratively contribute to constructing a global model without sharing their data. This is accomplished by devices computing local, private model updates that are then aggregated by a server. However, computational resource constraints and network communication can become a severe bottleneck for larger model sizes typical for deep learning applications. Edge nodes tend to have limited hardware resources (RAM, CPU), and the network bandwidth and reliability at the edge is a concern for scaling federated fleet applications. In this paper, we propose and evaluate a FL strategy inspired by transfer learning in order to reduce resource utilization on devices, as well as the load on the server and network in each global training round. For each local model update, we randomly select layers to train, freezing the remaining part of the model. In doing so, we can reduce both server load and communication costs per round by excluding all untrained layer weights from being transferred to the server. The goal of this study is to empirically explore the potential trade-off between resource utilization on devices and global model convergence under the proposed strategy. We implement the approach using the federated learning framework FEDn. A number of experiments were carried out over different datasets (CIFAR-10, CASA, and IMDB), performing different tasks using different deep-learning model architectures. Our results show that training the model partially can accelerate the training process, efficiently utilizes resources on-device, and reduce the data transmission by around 75% and 53% when we train 25%, and 50% of the model layers, respectively, without harming the resulting global model accuracy.

cs.LG↗

arXiv2023

FedBot: Enhancing Privacy in Chatbots with Federated Learning

Chatbots are mainly data-driven and usually based on utterances that might be sensitive. However, training deep learning models on shared data can violate user privacy. Such issues have commonly existed in chatbots since their inception. In the literature, there have been many approaches to deal with privacy, such as differential privacy and secure multi-party computation, but most of them need to have access to users' data. In this context, Federated Learning (FL) aims to protect data privacy through distributed learning methods that keep the data in its location. This paper presents Fedbot, a proof-of-concept (POC) privacy-preserving chatbot that leverages large-scale customer support data. The POC combines Deep Bidirectional Transformer models and federated learning algorithms to protect customer data privacy during collaborative model training. The results of the proof-of-concept showcase the potential for privacy-preserving chatbots to transform the customer support industry by delivering personalized and efficient customer service that meets data privacy regulations and legal requirements. Furthermore, the system is specifically designed to improve its performance and accuracy over time by leveraging its ability to learn from previous interactions.

cs.CL↗

arXiv2023

Accelerating Fair Federated Learning: Adaptive Federated Adam

Federated learning is a distributed and privacy-preserving approach to train a statistical model collaboratively from decentralized data of different parties. However, when datasets of participants are not independent and identically distributed (non-IID), models trained by naive federated algorithms may be biased towards certain participants, and model performance across participants is non-uniform. This is known as the fairness problem in federated learning. In this paper, we formulate fairness-controlled federated learning as a dynamical multi-objective optimization problem to ensure fair performance across all participants. To solve the problem efficiently, we study the convergence and bias of Adam as the server optimizer in federated learning, and propose Adaptive Federated Adam (AdaFedAdam) to accelerate fair federated learning with alleviated bias. We validated the effectiveness, Pareto optimality and robustness of AdaFedAdam in numerical experiments and show that AdaFedAdam outperforms existing algorithms, providing better convergence and fairness properties of the federated scheme.

cs.LG↗

arXiv2022

FedQAS: Privacy-aware machine reading comprehension with federated learning

Machine reading comprehension (MRC) of text data is one important task in Natural Language Understanding. It is a complex NLP problem with a lot of ongoing research fueled by the release of the Stanford Question Answering Dataset (SQuAD) and Conversational Question Answering (CoQA). It is considered to be an effort to teach computers how to "understand" a text, and then to be able to answer questions about it using deep learning. However, until now large-scale training on private text data and knowledge sharing has been missing for this NLP task. Hence, we present FedQAS, a privacy-preserving machine reading system capable of leveraging large-scale private data without the need to pool those datasets in a central location. The proposed approach combines transformer models and federated learning technologies. The system is developed using the FEDn framework and deployed as a proof-of-concept alliance initiative. FedQAS is flexible, language-agnostic, and allows intuitive participation and execution of local model training. In addition, we present the architecture and implementation of the system, as well as provide a reference evaluation based on the SQUAD dataset, to showcase how it overcomes data privacy issues and enables knowledge sharing between alliance members in a Federated learning setting.

cs.CL↗

arXiv2022

Efficient Hierarchical Storage Management Framework Empowered by Reinforcement Learning

With the rapid development of big data and cloud computing, data management has become increasingly challenging. Over the years, a number of frameworks for data management and storage with various characteristics and features have become available. Most of these are highly efficient, but ultimately create data silos. It becomes difficult to move and work coherently with data as new requirements emerge as no single framework can efficiently fulfill the data management needs of diverse applications. A possible solution is to design smart and efficient hierarchical (multi-tier) storage solutions. A hierarchical storage system (HSS) is a meta solution that consists of different storage frameworks organized as a jointly constructed large storage pool. It brings a number of benefits including better utilization of the storage, cost-efficiency, and use of different features provided by the underlying storage frameworks. In order to maximize the gains of hierarchical storage solutions, it is important that they include intelligent and autonomous mechanisms for data management grounded in the features of the different underlying frameworks. These decisions should be made according to the characteristics of the dataset, tier status, and access patterns. These are highly dynamic parameters and defining a policy based on the mentioned parameters is a non-trivial task. This paper presents an open-source hierarchical storage framework with a dynamic migration policy based on reinforcement learning (RL). We present a mathematical model, a software architecture, and an implementation based on both simulations and a live cloud-based environment. We compare the proposed RL-based strategy to a baseline of three rule-based policies, showing that the RL-based policy achieves significantly higher efficiency and optimal data distribution in different scenarios compared to the dynamic rule-based policies.

cs.DC↗

arXiv2021

Scalable federated machine learning with FEDn

Federated machine learning has great promise to overcome the input privacy challenge in machine learning. The appearance of several projects capable of simulating federated learning has led to a corresponding rapid progress on algorithmic aspects of the problem. However, there is still a lack of federated machine learning frameworks that focus on fundamental aspects such as scalability, robustness, security, and performance in a geographically distributed setting. To bridge this gap we have designed and developed the FEDn framework. A main feature of FEDn is to support both cross-device and cross-silo training settings. This makes FEDn a powerful tool for researching a wide range of machine learning applications in a realistic setting.

cs.LG↗

arXiv2021

Robust and integrative Bayesian neural networks for likelihood-free parameter inference

State-of-the-art neural network-based methods for learning summary statistics have delivered promising results for simulation-based likelihood-free parameter inference. Existing approaches require density estimation as a post-processing step building upon deterministic neural networks, and do not take network prediction uncertainty into account. This work proposes a robust integrated approach that learns summary statistics using Bayesian neural networks, and directly estimates the posterior density using categorical distributions. An adaptive sampling scheme selects simulation locations to efficiently and iteratively refine the predictive posterior of the network conditioned on observations. This allows for more efficient and robust convergence on comparatively large prior spaces. We demonstrate our approach on benchmark examples and compare against related methods.

stat.ML↗

arXiv2020

Convolutional Neural Networks as Summary Statistics for Approximate Bayesian Computation

Approximate Bayesian Computation is widely used in systems biology for inferring parameters in stochastic gene regulatory network models. Its performance hinges critically on the ability to summarize high-dimensional system responses such as time series into a few informative, low-dimensional summary statistics. The quality of those statistics acutely impacts the accuracy of the inference task. Existing methods to select the best subset out of a pool of candidate statistics do not scale well with large pools of several tens to hundreds of candidate statistics. Since high quality statistics are imperative for good performance, this becomes a serious bottleneck when performing inference on complex and high-dimensional problems. This paper proposes a convolutional neural network architecture for automatically learning informative summary statistics of temporal responses. We show that the proposed network can effectively circumvent the statistics selection problem of the preprocessing step for ABC inference. The proposed approach is demonstrated on two benchmark problem and one challenging inference problem learning parameters in a high-dimensional stochastic genetic oscillator. We also study the impact of experimental design on network performance by comparing different data richness and data acquisition strategies.

stat.ML↗

arXiv2020

Smart Resource Management for Data Streaming using an Online Bin-packing Strategy

Data stream processing frameworks provide reliable and efficient mechanisms for executing complex workflows over large datasets. A common challenge for the majority of currently available streaming frameworks is efficient utilization of resources. Most frameworks use static or semi-static settings for resource utilization that work well for established use cases but lead to marginal improvements for unseen scenarios. Another pressing issue is the efficient processing of large individual objects such as images and matrices typical for scientific datasets. HarmonicIO has proven to be a good solution for streams of relatively large individual objects, as demonstrated in a benchmark comparison with the Spark and Kafka streaming frameworks. We here present an extension of the HarmonicIO framework based on the online bin-packing algorithm, to allow for efficient utilization of resources. Based on a real world use case from large-scale microscopy pipelines, we compare results of the new system to Spark's auto-scaling mechanism.

cs.DC↗

arXiv2019

Apache Spark Streaming, Kafka and HarmonicIO: A Performance Benchmark and Architecture Comparison for Enterprise and Scientific Computing

This paper presents a benchmark of stream processing throughput comparing Apache Spark Streaming (under file-, TCP socket- and Kafka-based stream integration), with a prototype P2P stream processing framework, HarmonicIO. Maximum throughput for a spectrum of stream processing loads are measured, specifically, those with large message sizes (up to 10MB), and heavy CPU loads -- more typical of scientific computing use cases (such as microscopy), than enterprise contexts. A detailed exploration of the performance characteristics with these streaming sources, under varying loads, reveals an interplay of performance trade-offs, uncovering the boundaries of good performance for each framework and streaming source integration. We compare with theoretic bounds in each case. Based on these results, we suggest which frameworks and streaming sources are likely to offer good performance for a given load. Broadly, the advantages of Spark's rich feature set comes at a cost of sensitivity to message size in particular -- common stream source integrations can perform poorly in the 1MB-10MB range. The simplicity of HarmonicIO offers more robust performance in this region, especially for raw CPU utilization.

cs.DC↗

arXiv2019

Resource- and Message Size-Aware Scheduling of Stream Processing at the Edge with application to Realtime Microscopy

Whilst computational resources at the cloud edge can be leveraged to improve latency and reduce the costs of cloud services for a wide variety mobile, web, and IoT applications; such resources are naturally constrained. For distributed stream processing applications, there are clear advantages to offloading some processing work to the cloud edge. Many state of the art stream processing applications such as Flink and Spark Streaming, being designed to run exclusively in the cloud, are a poor fit for such hybrid edge/cloud deployment settings, not least because their schedulers take limited consideration of the heterogeneous hardware in such deployments. In particular, their schedulers broadly assume a homogeneous network topology (aside from data locality consideration in, e.g., HDFS/Spark). Specialized stream processing frameworks intended for such hybrid deployment scenarios, especially IoT applications, allow developers to manually allocate specific operators in the pipeline to nodes at the cloud edge. In this paper, we investigate scheduling stream processing in hybrid cloud/edge deployment settings with sensitivity to CPU costs and message size, with the aim of maximizing throughput with respect to limited edge resources. We demonstrate real-time edge processing of a stream of electron microscopy images, and measure a consistent reduction in end-to-end latency under our approach versus a resource-agnostic baseline scheduler, under benchmarking.

cs.DC↗