arXiv subjects

Marco Morik

Publications and source records attributed to Marco Morik.

9 recordsLinked to original sources

arXiv2026

Structure-Preserving Correction Learning for Sparse Bayesian Inference in Brain Source Imaging

Classical sparse Type-II Bayesian methods for M/EEG brain imaging support joint estimation of source and noise hyperparameters, but rely on fixed iterative update rules. Although these updates are principled and interpretable, their dynamics cannot be adapted from data. We propose to learn the update mechanism itself while preserving the underlying Bayesian structure by unfolding a classical joint hyperparameter-learning solver into a trainable neural architecture whose layers mirror the original iterations. The resulting framework is initialized to recover the classical solver exactly before training and is enriched through progressively more expressive correction-learning mechanisms, ranging from learnable biases to adaptive MLP and attention-based contextual refinements. In this way, training does not replace Bayesian inference with a black-box predictor, but instead learns structured correction terms while retaining the interpretability and model-based character of the original update dynamics. Structured correction learning therefore aims to improve empirical reconstruction performance without replacing the original model-based inference mechanism. Experimental results show that the learned correction variants improve reconstruction performance and convergence behavior over the baseline unfolded solver while preserving its algorithmic transparency.

cs.LG↗

arXiv2026

Visual Disentangled Diffusion Autoencoders: Scalable Counterfactual Generation for Foundation Models

Foundation models, despite their robust zero-shot capabilities, remain vulnerable to spurious correlations and 'Clever Hans' strategies. Existing mitigation methods often rely on unavailable group labels or computationally expensive gradient-based adversarial optimization. To address these limitations, we propose Visual Disentangled Diffusion Autoencoders (DiDAE), a novel framework integrating frozen foundation models with disentangled dictionary learning for efficient, gradient-free counterfactual generation directly for the foundation model. DiDAE first edits foundation model embeddings in interpretable disentangled directions of the disentangled dictionary and then decodes them via a diffusion autoencoder. This allows the generation of multiple diverse, disentangled counterfactuals for each factual, much faster than existing baselines, which generate single entangled counterfactuals. When paired with Counterfactual Knowledge Distillation, DiDAE-CFKD achieves state-of-the-art performance in mitigating shortcut learning, improving downstream performance on unbalanced datasets.

cs.LG↗

arXiv2026

Attentive multilayer fusion for vision transformers

With the rise of large-scale foundation models, efficiently adapting them to downstream tasks remains a central challenge. Linear probing, which freezes the backbone and trains a lightweight head, is computationally efficient but often restricted to last-layer representations. We show that task-relevant information is distributed across the network hierarchy rather than encoded solely in the last layers. To leverage this distribution of information, we apply an attentive probing mechanism that dynamically fuses representations from all layers of a Vision Transformer. This attentive layer fusion (ALF) learns to identify the most relevant layers for a target task and combines low-level structural cues with high-level semantic abstractions. Across 20 diverse datasets and multiple pretrained foundation models, ALF achieves consistent, substantial gains over standard linear probes. Attention heatmaps further reveal that tasks different from the pre-training domain benefit most from intermediate representations. Overall, our findings underscore the value of intermediate layers and demonstrate a principled, task-aware approach for unlocking their potential for probing-based adaptation.

cs.CV↗

arXiv2024

Objective drives the consistency of representational similarity across datasets

The Platonic Representation Hypothesis claims that recent foundation models are converging to a shared representation space as a function of their downstream task performance, irrespective of the objectives and data modalities used to train these models (Huh et al., 2024). Representational similarity is generally measured for individual datasets and is not necessarily consistent across datasets. Thus, one may wonder whether this convergence of model representations is confounded by the datasets commonly used in machine learning. Here, we propose a systematic way to measure how representational similarity between models varies with the set of stimuli used to construct the representations. We find that the objective function is a crucial factor in determining the consistency of representational similarities across datasets. Specifically, self-supervised vision models learn representations whose relative pairwise similarities generalize better from one dataset to another compared to those of image classification or image-text models. Moreover, the correspondence between representational similarities and the models' task behavior is dataset-dependent, being most strongly pronounced for single-domain datasets. Our work provides a framework for analyzing similarities of model representations across datasets and linking those similarities to differences in task behavior.

cs.CV↗

arXiv2024

Enhancing Brain Source Reconstruction by Initializing 3D Neural Networks with Physical Inverse Solutions

Reconstructing brain sources is a fundamental challenge in neuroscience, crucial for understanding brain function and dysfunction. Electroencephalography (EEG) signals have a high temporal resolution. However, identifying the correct spatial location of brain sources from these signals remains difficult due to the ill-posed structure of the problem. Traditional methods predominantly rely on manually crafted priors, missing the flexibility of data-driven learning, while recent deep learning approaches focus on end-to-end learning, typically using the physical information of the forward model only for generating training data. We propose the novel hybrid method 3D-PIUNet for EEG source localization that effectively integrates the strengths of traditional and deep learning techniques. 3D-PIUNet starts from an initial physics-informed estimate by using the pseudo inverse to map from measurements to source space. Secondly, by viewing the brain as a 3D volume, we use a 3D convolutional U-Net to capture spatial dependencies and refine the solution according to the learned data prior. Training the model relies on simulated pseudo-realistic brain source data, covering different source distributions. Trained on this data, our model significantly improves spatial accuracy, demonstrating superior performance over both traditional and end-to-end data-driven methods. Additionally, we validate our findings with real EEG data from a visual task, where 3D-PIUNet successfully identifies the visual cortex and reconstructs the expected temporal behavior, thereby showcasing its practical applicability.

eess.IV↗

arXiv2024

An Analysis of Human Alignment of Latent Diffusion Models

Diffusion models, trained on large amounts of data, showed remarkable performance for image synthesis. They have high error consistency with humans and low texture bias when used for classification. Furthermore, prior work demonstrated the decomposability of their bottleneck layer representations into semantic directions. In this work, we analyze how well such representations are aligned to human responses on a triplet odd-one-out task. We find that despite the aforementioned observations: I) The representational alignment with humans is comparable to that of models trained only on ImageNet-1k. II) The most aligned layers of the denoiser U-Net are intermediate layers and not the bottleneck. III) Text conditioning greatly improves alignment at high noise levels, hinting at the importance of abstract textual information, especially in the early stage of generation.

cs.LG↗

arXiv2020

Controlling Fairness and Bias in Dynamic Learning-to-Rank

Rankings are the primary interface through which many online platforms match users to items (e.g. news, products, music, video). In these two-sided markets, not only the users draw utility from the rankings, but the rankings also determine the utility (e.g. exposure, revenue) for the item providers (e.g. publishers, sellers, artists, studios). It has already been noted that myopically optimizing utility to the users, as done by virtually all learning-to-rank algorithms, can be unfair to the item providers. We, therefore, present a learning-to-rank approach for explicitly enforcing merit-based fairness guarantees to groups of items (e.g. articles by the same publisher, tracks by the same artist). In particular, we propose a learning algorithm that ensures notions of amortized group fairness, while simultaneously learning the ranking function from implicit feedback data. The algorithm takes the form of a controller that integrates unbiased estimators for both fairness and utility, dynamically adapting both as more data becomes available. In addition to its rigorous theoretical foundation and convergence guarantees, we find empirically that the algorithm is highly practical and robust.

cs.IR↗

arXiv2019

Listing All Maximal $k$-Plexes in Temporal Graphs

Many real-world networks evolve over time, that is, new contacts appear and old contacts may disappear. They can be modeled as temporal graphs where interactions between vertices (which represent people in the case of social networks) are represented by time-stamped edges. One of the most fundamental problems in (social) network analysis is community detection, and one of the most basic primitives to model a community is a clique. Addressing the problem of finding communities in temporal networks, Viard et al. [TCS 2016] introduced $Δ$-cliques as a natural temporal version of cliques. Himmel et al. [SNAM 2017] showed how to adapt the well-known Bron-Kerbosch algorithm to enumerate $Δ$-cliques. We continue this work and improve and extend the algorithm of Himmel et al. to enumerate temporal $k$-plexes (notably, cliques are the special case $k=1$). We define a $Δ$-$k$-plex as a set of vertices and a time interval, where during this time interval each vertex has in each consecutive $Δ+ 1$ time steps at least one edge to all but at most $k-1$ vertices in the chosen set of vertices. We develop a recursive algorithm for enumerating all maximal $Δ$-$k$-plexs and perform experiments on real-world social networks that demonstrate the practical feasibility of our approach. In particular, for the special case of $Δ$-1-plexes (that is, $Δ$-cliques), we observe that our algorithm is on average significantly faster than the previous algorithms by Himmel et al. [SNAM 2017] and Viard et al. [IPL 2018] for enumerating $Δ$-cliques.

cs.DS↗

arXiv2017

The Complexity of Routing with Few Collisions

We study the computational complexity of routing multiple objects through a network in such a way that only few collisions occur: Given a graph $G$ with two distinct terminal vertices and two positive integers $p$ and $k$, the question is whether one can connect the terminals by at least $p$ routes (e.g. paths) such that at most $k$ edges are time-wise shared among them. We study three types of routes: traverse each vertex at most once (paths), each edge at most once (trails), or no such restrictions (walks). We prove that for paths and trails the problem is NP-complete on undirected and directed graphs even if $k$ is constant or the maximum vertex degree in the input graph is constant. For walks, however, it is solvable in polynomial time on undirected graphs for arbitrary $k$ and on directed graphs if $k$ is constant. We additionally study for all route types a variant of the problem where the maximum length of a route is restricted by some given upper bound. We prove that this length-restricted variant has the same complexity classification with respect to paths and trails, but for walks it becomes NP-complete on undirected graphs.

cs.CC↗