arXiv subjects

Kristian Hovde Liland

Publications and source records attributed to Kristian Hovde Liland.

10 recordsLinked to original sources

arXiv2026

Continual Test-Time Adaptation via Entropy Sensitivity-Guidance in Strict Online Setting

Test-time adaptation (TTA) promises robustness under distribution shift by updating a pretrained model on unlabeled test data, but strict online TTA with batch size one and no access to source data is especially prone to drift or collapse. We introduce Sensitivity-Guided Erasing Adaptation (SEGA), a method for strict online continual TTA (CTTA) on corruption-style streams. SEGA uses a small number of structured erasures to probe how predictive entropy changes as information is removed, and uses the resulting per-sample sensitivity trajectories to coordinate recovery and sample selection rather than relying on raw entropy or batch statistics. This yields a practical feedback signal for long-horizon batch-size-one adaptation without periodic resets or model reservoirs. In experiments on ImageNet-C, CIFAR10/100-C, and corruption-generated aquaculture streams treated as controlled corruption-style proxies, SEGA yields consistent robustness and stability gains over strong CTTA baselines while reducing backward passes through sensitivity-based gating.

cs.CV↗

arXiv2026

Towards Continual Test-Time Adaptation of Vision-Language Models in Open-Vocabulary Semantic Segmentation

Open-vocabulary semantic segmentation (OVSS) relies on vision-language alignment to recognize arbitrary text-defined categories, yet this alignment is fragile under continual test-time distribution shift. Our diagnostic analysis reveals that entropy minimization drives patch-level class collapse, continual updates erode vision-language alignment, and redundant gradients from low-shift samples waste computation. We propose Diversify, Anchor, and Filter (DAF), a stabilization framework that augments entropy-based adaptation with a marginal diversity loss that resists collapse, a cross-modal anchor consistency loss that constrains feature drift relative to a frozen source model, and feature salience filtering that skips low-value backward passes to offset part of the source-anchor overhead. We evaluate on five datasets spanning natural scenes, autonomous driving, underwater imagery, and remote sensing with their corrupted variants. Across the evaluated continual shifts, DAF remains stable where entropy minimization collapses, improving mIoU by over 8 points on Pascal VOC20-C, over 9 points on LoveDA, and over 3 points on Foggy Cityscapes compared to the source model, and is robust to aggressive adaptation and learning rate choices.

cs.CV↗

arXiv2026

Family Matters: A Systematic Study of Spatial vs. Frequency Masking for Continual Test-Time Adaptation

Recent continual test-time adaptation (CTTA) methods adopt masked image modeling to stabilize learning under distribution shift, yet each treats its masking family F as a fixed design choice and innovates exclusively along the selection strategy S, leaving the family axis underexplored. We present a systematic empirical study that isolates this axis. Using a controlled CTTA instantiation, Mask to Adapt (M2A), that fixes S=random and standard losses, we vary only F across spatial (patch, pixel) and frequency (all-band, low-band, high-band) families while keeping every other component identical. The study's contributions are the design guidance it extracts for the CTTA settings we evaluated: (1) the masking family determines whether adaptation compounds useful structure or compounds errors, on patch-tokenized architectures, spatial masking accumulates stable representations over long streams while frequency masking collapses catastrophically. We characterize this instability through a structural-preservation account, where spatial coherence maintains the broad-spectrum redundancy needed to avoid terminally overlapping with a corruption's spectral signature; (2) the optimal family depends on architecture-task alignment, on CNNs, whose overlapping receptive fields dilute patch occlusion, the family gap vanishes, whereas on fine-grained tasks with global cues and large-capacity ViTs, frequency masking becomes competitive. In confounded system-level comparisons, where baselines also differ in losses and auxiliary components, M2A's random selection performs comparably to heuristic strategies, though we treat this observation as suggestive context rather than a controlled quantification of S's relative importance.

cs.CV↗

arXiv2026

Towards Sustainable Universal Deepfake Detection with Frequency-Domain Masking

Universal deepfake detection aims to identify AI-generated images across a broad range of generative models, including unseen ones. This requires robust generalization to new and unseen deepfakes, which emerge frequently, while minimizing computational overhead to enable large-scale deepfake screening, a critical objective in the era of Green AI. In this work, we explore frequency-domain masking as a training strategy for deepfake detectors. Unlike traditional methods that rely heavily on spatial features or large-scale pretrained models, our approach introduces random masking and geometric transformations, with a focus on frequency masking due to its superior generalization properties. We demonstrate that frequency masking not only enhances detection accuracy across diverse generators but also maintains performance under significant model pruning, offering a scalable and resource-conscious solution. Our method achieves state-of-the-art generalization on GAN- and diffusion-generated image datasets and exhibits consistent robustness under structured pruning. These results highlight the potential of frequency-based masking as a practical step toward sustainable and generalizable deepfake detection. Code and models are available at https://github.com/chandlerbing65nm/FakeImageDetection.

cs.CV↗

arXiv2024

General Effect Modelling (GEM) -- Part 1. Method description

We present a flexible tool, called General Effect Modelling (GEM), for the analysis of any multivariate data influenced by one or more qualitative (categorical) or quantitative (continuous) input variables. The variables can be design factors or observed values, e.g., age, sex, or income, or they may represent subgroups of the samples found by data exploration. The first step in GEM separates the variation in the multivariate data into effect matrices associated with each of the influencing variables (and possibly interactions between these) by applying a general linear model. The residuals of the model are added to each of the effect matrices and the results are called Effect plus Residual (ER) values. The tables of ER values have the same dimensions as the original multivariate data. The second step of GEM is a multi- or univariate exploration of the ER tables to learn more about the multivariate data in relation to each input variables. The exploration is simplified as it addresses one input variable at the time, or if preferred, a combination of input variables. One example is a study to identify molecular fingerprints associated with a disease that is not influenced by age where individuals at different ages with and without the disease are included in the experiment. This situation can be described as an experiment with two input variables: the targeted disease and the individual age. Through GEM, the effect of age can be removed, thus focusing further analysis on the targeted disease without the influence of the confounding effect of age. ER values can also be the combined effect of several input variables. This publication has three parts: the first part describes the GEM methodology, Part 2 is a consideration of multivariate data and the benefits of treating such data by multivariate analysis, with a focus on omics data, and Part 3 is a case study in Multiple Sclerosis (MS).

stat.ME↗

arXiv2024

General Effect Modelling (GEM) -- Part 2. Multivariate GEM applied to gene expression data of type 2 diabetes detects information that is lost by univariate validation

General Effect Modelling (GEM) is an umbrella over different methods that utilise effects in the analyses of data with multiple design variables and multivariate responses. To demonstrate the methodology, we here use GEM in gene expression data where we use GEM to combine data from different cohorts and apply multivariate analysis of the effects of the targeted disease across the cohorts. Omics data are by nature multivariate, yet univariate analysis is the dominating approach used for such data. A major challenge in omics data is that the number of features such as genes, proteins and metabolites are often very large, whereas the number of samples is limited. Furthermore, omics research aims to obtain results that are generically valid across different backgrounds. The present publication applies GEM to address these aspects. First, we emphasise the benefit of multivariate analysis for multivariate data. Then we illustrate the use of GEM to combine data from two different cohorts for multivariate analysis across the cohorts, and we highlight that multivariate analysis can detect information that is lost by univariate validation.

stat.ME↗

arXiv2024

General Effect Modelling (GEM) -- Part 3. GEM applied on proteome data of cerebrospinal fluid of multiple sclerosis and clinically isolated syndrome

The novel data analytical platform General Effect Modelling (GEM), is an umbrella platform covering different data analytical methods that handle data with multiple design variables (or pseudo design variables) and multivariate responses. GEM is here demonstrated in an analysis of proteome data from cerebrospinal fluid (CSF) from two independent previously published datasets, one data set comprised of persons with relapsing-remitting multiple sclerosis, persons with other neurological disorders and persons without neurological disorders, and one data set had persons with clinically isolated syndrome (CIS), which is the first clinical symptom of MS, and controls. The primary aim of the present publication is to use these data to demonstrate how patient stratification can be utilised by GEM for multivariate analysis. We also emphasize how the findings shed light on important aspects of the molecular mechanism of MS that may otherwise be lost. We identified proteins involved in neural development as significantly lower for MS/CIS than for their respective controls. This information was only seen after stratification of the persons into two groups, which were found to have different inflammatory patterns and the utilisation of this by GEM. Our conclusion from the study of these data is that disrupted neural development may be an early event in CIS and MS.

stat.ME↗

arXiv2024

A New Formula for Faster Computation of the K-Fold Cross-Validation and Good Regularisation Parameter Values in Ridge Regression

In the present paper, we prove a new theorem, resulting in an update formula for linear regression model residuals calculating the exact k-fold cross-validation residuals for any choice of cross-validation strategy without model refitting. The required matrix inversions are limited by the cross-validation segment sizes and can be executed with high efficiency in parallel. The well-known formula for leave-one-out cross-validation follows as a special case of the theorem. In situations where the cross-validation segments consist of small groups of repeated measurements, we suggest a heuristic strategy for fast serial approximations of the cross-validated residuals and associated Predicted Residual Sum of Squares (PRESS) statistic. We also suggest strategies for efficient estimation of the minimum PRESS value and full PRESS function over a selected interval of regularisation values. The computational effectiveness of the parameter selection for Ridge- and Tikhonov regression modelling resulting from our theoretical findings and heuristic arguments is demonstrated in several applications with real and highly multivariate datasets.

stat.ME↗

arXiv2022

Ranking Feature-Block Importance in Artificial Multiblock Neural Networks

In artificial neural networks, understanding the contributions of input features on the prediction fosters model explainability and delivers relevant information about the dataset. While typical setups for feature importance ranking assess input features individually, in this study, we go one step further and rank the importance of groups of features, denoted as feature-blocks. A feature-block can contain features of a specific type or features derived from a particular source, which are presented to the neural network in separate input branches (multiblock ANNs). This work presents three methods pursuing distinct strategies to rank features in multiblock ANNs by their importance: (1) a composite strategy building on individual feature importance rankings, (2) a knock-in, and (3) a knock-out strategy. While the composite strategy builds on state-of-the-art feature importance rankings, knock-in and knock-out strategies evaluate the block as a whole via a mutual information criterion. Our experiments consist of a simulation study validating all three approaches, followed by a case study on two distinct real-world datasets to compare the strategies. We conclude that each strategy has its merits for specific application scenarios.

cs.LG↗

arXiv2021

RENT -- Repeated Elastic Net Technique for Feature Selection

Feature selection is an essential step in data science pipelines to reduce the complexity associated with large datasets. While much research on this topic focuses on optimizing predictive performance, few studies investigate stability in the context of the feature selection process. In this study, we present the Repeated Elastic Net Technique (RENT) for Feature Selection. RENT uses an ensemble of generalized linear models with elastic net regularization, each trained on distinct subsets of the training data. The feature selection is based on three criteria evaluating the weight distributions of features across all elementary models. This fact leads to the selection of features with high stability that improve the robustness of the final model. Furthermore, unlike established feature selectors, RENT provides valuable information for model interpretation concerning the identification of objects in the data that are difficult to predict during training. In our experiments, we benchmark RENT against six established feature selectors on eight multivariate datasets for binary classification and regression. In the experimental comparison, RENT shows a well-balanced trade-off between predictive performance and stability. Finally, we underline the additional interpretational value of RENT with an exploratory post-hoc analysis of a healthcare dataset.

cs.LG↗