Clustered Federated Learning has emerged as an effective approach for handling heterogeneous data across clients by partitioning them into clusters with similar or identical data distributions. However, most existing methods, including the Iterative Federated Clustering Algorithm (IFCA), rely on a central server to coordinate model updates, typically requiring stable connectivity, synchronous communication rounds, and global aggregation of client models. These assumptions are difficult to satisfy in decentralized and heterogeneous environments, where clients may only have limited, local communication with a small subset of peers. As a result, such methods create a bottleneck and a single point of failure, limiting their applicability in realistic decentralized learning settings. This limitation is particularly severe in Internet of Things settings, where large numbers of resource-constrained devices, intermittent or sparse connectivity, and dynamic participation make reliance on a central server impractical. In this work, we introduce the Decentralized Federated Clustering Algorithm (DFCA), a fully decentralized clustered federated learning algorithm that enables clients to collaboratively train cluster-specific models without central coordination. DFCA uses a sequential running average to aggregate models from neighbors as updates arrive, providing a communication-efficient alternative to batch aggregation while maintaining clustering performance. Our experiments on various datasets demonstrate that DFCA outperforms other decentralized algorithms and performs comparably to centralized IFCA, even under sparse connectivity, highlighting its robustness and practicality for dynamic real-world decentralized networks.

What are the essential components of world models? How far do we get with world models that are not employing RNNs, transformers, discrete representations, and image reconstructions? This paper introduces SGF, a Simple, Good, and Fast world model that uses self-supervised representation learning, captures short-time dependencies through frame and action stacking, and enhances robustness against model errors through data augmentation. We extensively discuss SGF’s connections to established world models, evaluate the building blocks in ablation studies, and demonstrate good performance through quantitative comparisons on the Atari 100k benchmark. The code is available at https://github.com/jrobine/sgf.

Stance detection holds great potential to improve online political discussions through its deployment in discussion platforms for purposes such as content moderation, topic summarisation or to facilitate more balanced discussions. Typically, transformer-based models are employed directly for stance detection, requiring vast amounts of data. However, the wide variety of debate topics in online political discussions makes data collection particularly challenging. LLMs have revived stance detection, but their online deployment in online political discussions faces challenges like inconsistent outputs, biases, and vulnerability to adversarial attacks. We show how LLM-generated synthetic data can improve stance detection for online political discussions by using reliable traditional stance detection models for online deployment, while leveraging the text generation capabilities of LLMs for synthetic data generation in a secure offline environment. To achieve this, (i) we generate synthetic data for specific debate questions by prompting a Mistral-7B model and show that fine-tuning with the generated synthetic data can substantially improve the performance of stance detection, while remaining interpretable and aligned with real world data. (ii) Using the synthetic data as a reference, we can improve performance even further by identifying the most informative samples in an unlabelled dataset, i.e., those samples which the stance detection model is most uncertain about and can benefit from the most. By fine-tuning with both synthetic data and the most informative samples, we surpass the performance of the baseline model that is fine-tuned on all true labels, while labelling considerably less data.

Abstract Learning meaningful representations is at the heart of many tasks in the field of modern machine learning. Recently, a lot of methods were introduced that allow learning of image representations without supervision. These representations can then be used in downstream tasks like classification or object detection. The quality of these representations is close to supervised learning, while no labeled images are needed. This survey paper provides a comprehensive review of these methods in a unified notation, points out similarities and differences of these methods, and proposes a taxonomy which sets these methods in relation to each other. Furthermore, our survey summarizes the most recent experimental results reported in the literature in form of a meta-study. Our survey is intended as a starting point for researchers and practitioners who want to dive into the field of representation learning.

In this paper we adopt a representation-centric perspective on exploration in reinforcement learning, viewing exploration fundamentally as a density estimation problem. We investigate the effectiveness of clustering representations for exploration in 3-D environments, based on the observation that the importance of pixel changes between transitions is less pronounced in 3-D environments compared to 2-D environments, where pixel changes between transitions are typically distinct and significant. We propose a method that performs episodic and global clustering on random representations and on pre-trained DINO representations to count states, i.e, estimate pseudo-counts. Surprisingly, even random features can be clustered effectively to count states in 3-D environments, however when these become visually more complex, pre-trained DINO representations are more effective thanks to the pre-trained inductive biases in the representations. Overall, this presents a pathway for integrating pre-trained biases into exploration. We evaluate our approach on the VizDoom and Habitat environments, demonstrating that our method surpasses other well-known exploration methods in these settings.

Reconstructing images from magnitude measurements is an important and difficult problem arising in many research areas, such as X-ray crystallography, astronomical imaging and more. While optimization-based approaches often struggle with the non-convexity and non- linearity of the problem, learning-based approaches are able to produce reconstructions of high quality for data similar to a given training dataset. In this work, we analyze a class of methods based on conditional generative adversarial networks (CGAN). We show how the benefits of optimization-based and learning-based methods can be combined to improve reconstruction quality. Furthermore, we show that these combined methods are able to generalize to out-of-distribution data and analyze their robustness to measurement noise. In addition to that, we compare how the methods are impacted by missing measurements. Extensive ablation studies demonstrate that all components of our approach are essential and justify the choice of network architecture.

Abstract Model-based reinforcement learning algorithms try to learn an agent by training a model that simulates the environment. However, the size of such models tends to be quite large which could be a burden as well. In this paper, we address the question, how we could design a model with fewer parameters than previous model-based approaches while achieving the same performance in the 100 K-interactions regime. For this purpose, we create a world model that combines a vector quantized-variational autoencoder to encode observations and a convolutional long short-term memory to model the dynamics. This is connected to a model-free proximal policy optimization agent to train purely on simulated experience from this world model. Detailed experiments on the Atari environments show that it is possible to reach comparable performance to the SimPLe method with a significantly smaller world model. A series of ablation studies justify our design choices and give additional insights.

We present a conceptually simple framework for object instance segmentation, called Contour Proposal Network (CPN), which detects possibly overlapping objects in an image while simultaneously fitting closed object contours using a fixed-size representation based on Fourier Descriptors. The CPN can incorporate state-of-the-art object detection architectures as backbone networks into a single-stage instance segmentation model that can be trained end-to-end. We construct CPN models with different backbone networks and apply them to instance segmentation of cells in datasets from different modalities. In our experiments, CPNs outperform U-Net, Mask R-CNN and StarDist in instance segmentation accuracy. We present variants with execution times suitable for real-time applications. The trained models generalize well across different domains of cell types. Since the main assumption of the framework is closed object contours, it is applicable to a wide range of detection problems also beyond the biomedical domain. An implementation of the model architecture in PyTorch is freely available.

Exchanging arguments is an important part in communication, but we are often flooded with lots of arguments for different positions or are captured in filter bubbles.Tools which can present strong arguments relevant to oneself could help to reduce those problems.To be able to evaluate algorithms which can predict how convincing an argument is, we have collected a dataset with more than 900 arguments and personal attitudes of 600 individuals, which we present in this paper.Based on this data, we suggest three recommender tasks, for which we provide two baseline results from a simple majority classifier and a more complex nearest-neighbor algorithm.Our results suggest that better algorithms can still be developed, and we invite the community to improve on our results.

We propose a modular extension of backpropagation for the computation of block-diagonal approximations to various curvature matrices of the training objective (in particular, the Hessian, generalized Gauss-Newton, and positive-curvature Hessian). The approach reduces the otherwise tedious manual derivation of these matrices into local modules, and is easy to integrate into existing machine learning libraries. Moreover, we develop a compact notation derived from matrix differential calculus. We outline different strategies applicable to our method. They subsume recently-proposed block-diagonal approximations as special cases, and are extended to convolutional neural networks in this work.

Alpha matting aims to estimate the translucency of an object in a given image. The resulting alpha matte describes pixel-wise to what amount foreground and background colors contribute to the color of the composite image. While most methods in literature focus on estimating the alpha matte, the process of estimating the foreground colors given the input image and its alpha matte is often neglected, although foreground estimation is an essential part of many image editing workflows. In this work, we propose a novel method for foreground estimation given the alpha matte. We demonstrate that our fast multi-level approach yields results that are comparable with the state-of-the-art while outperforming those methods in computational runtime and memory usage.

In this paper, we propose the application of conditional generative adversarial networks to solve various phase retrieval problems. We show that including knowledge of the measurement process at training time leads to an optimization at test time that is more robust to initialization than existing approaches involving generative models. In addition, conditioning the generator network on the measurements enables us to achieve much more detailed results. We empirically demonstrate that these advantages provide meaningful solutions to the Fourier and the compressive phase retrieval problem and that our method outperforms well-established projection-based methods as well as existing methods that are based on neural networks. Like other deep learning methods, our approach is robust to noise and can therefore be useful for real-world applications.

We describe a learning-based approach to blind image deconvolution. It uses a deep layered architecture, parts of which are borrowed from recent work on neural network learning, and parts of which incorporate computations that are specific to image deconvolution. The system is trained end-to-end on a set of artificially generated training examples, enabling competitive performance in blind deconvolution, both with respect to quality and runtime.

We study the problem of object recognition for categories for which we have no training examples, a task also called zero--data or zero-shot learning. This situation has hardly been studied in computer vision research, even though it occurs frequently; the world contains tens of thousands of different object classes, and image collections have been formed and suitably annotated for only a few of them. To tackle the problem, we introduce attribute-based classification: Objects are identified based on a high-level description that is phrased in terms of semantic attributes, such as the object's color or shape. Because the identification of each such property transcends the specific learning task at hand, the attribute classifiers can be prelearned independently, for example, from existing image data sets unrelated to the current task. Afterward, new classes can be detected based on their attribute representation, without the need for a new training phase. In this paper, we also introduce a new data set, Animals with Attributes, of over 30,000 images of 50 animal classes, annotated with 85 semantic attributes. Extensive experiments on this and two more data sets show that attribute-based classification indeed is able to categorize images without access to any training images of the target classes.

We establish a link between Fourier optics and a recent construction from the machine learning community termed the kernel mean map. Using the Fraunhofer approximation, it identifies the kernel with the squared Fourier transform of the aperture. This allows us to use results about the invertibility of the kernel mean map to provide a statement about the invertibility of Fraunhofer diffraction, showing that imaging processes with arbitrarily small apertures can in principle be invertible, i.e., do not lose information, provided the objects to be imaged satisfy a generic condition. A real world experiment shows that we can super-resolve beyond the Rayleigh limit.

Image deconvolution is the ill-posed problem of recovering a sharp image, given a blurry one generated by a convolution. In this work, we deal with space-invariant non-blind deconvolution. Currently, the most successful methods involve a regularized inversion of the blur in Fourier domain as a first step. This step amplifies and colors the noise, and corrupts the image information. In a second (and arguably more difficult) step, one then needs to remove the colored noise, typically using a cleverly engineered algorithm. However, the methods based on this two-step approach do not properly address the fact that the image information has been corrupted. In this work, we also rely on a two-step procedure, but learn the second step on a large dataset of natural images, using a neural network. We will show that this approach outperforms the current state-of-the-art on a large dataset of artificially blurred images. We demonstrate the practical applicability of our method in a real-world example with photographic out-of-focus blur.

We present a new method for separating motion blurred foreground objects from their background given a single image. Previous techniques focused on estimating alpha mattes for separating sharp, non-moving foreground objects from fairly homogeneous background. In those cases the only pixels which are ambiguous are those which exhibit fractional pixel occupancy. In this paper, we address the problem of alpha matte and foreground estimation of motion blurred objects. We show, that explicit modeling of the object motion facilitates the estimation and improves the quality of the estimated alpha mattes. In addition, we improve foreground extraction of motion blurred objects with a new regularization term. This task is particularly difficult in smeared out regions, where the background shimmers through. Both synthetic and real-world examples illustrate the merit of our approach.

Image denoising can be described as the problem of mapping from a noisy image to a noise-free image. The best currently available denoising methods approximate this mapping with cleverly engineered algorithms. In this work we attempt to learn this mapping directly with a plain multi layer perceptron (MLP) applied to image patches. While this has been done before, we will show that by training on large image databases we are able to compete with the current state-of-the-art image denoising methods. Furthermore, our approach is easily adapted to less extensively studied types of noise (by merely exchanging the training data), for which we achieve excellent results as well.

For digital photographs of astronomical objects, where exposure times are usually long and ISO settings high, the so-called dark-current is a significant source of noise. Dark-current refers to thermally generated electrons and is therefore present even in the absence of light. This paper presents a novel approach for denoising astronomical images that have been corrupted by dark-current noise. Our method relies on a probabilistic description of the dark-current of each pixel of a given camera. The noise model is then combined with an image prior which is adapted to astronomical images. In a laboratory environment, we use a black and white CCD camera containing a cooling unit and show that our method is superior to existing methods in terms of root mean squared error. Furthermore, we show that our method is practically relevant by providing visually more appealing results on astronomical photographs taken with a single lens reflex CMOS camera.

Camera shake leads to non-uniform image blurs. State-of-the-art methods for removing camera shake model the blur as a linear combination of homographically transformed versions of the true image. While this is conceptually interesting, the resulting algorithms are computationally demanding. In this paper we develop a forward model based on the efficient filter flow framework, incorporating the particularities of camera shake, and show how an efficient algorithm for blur removal can be obtained. Comprehensive comparisons on a number of real-world blurry images show that our approach is not only substantially faster, but it also leads to better deblurring results.

Taking a sharp photo at several megapixel resolution traditionally relies on high grade lenses. In this paper, we present an approach to alleviate image degradations caused by imperfect optics. We rely on a calibration step to encode the optical aberrations in a space-variant point spread function and obtain a corrected image by non-stationary deconvolution. By including the Bayer array in our image formation model, we can perform demosaicing as part of the deconvolution.

A challenging problem in image restoration is to recover an image with a blurry foreground. Such images can easily occur with modern cameras, when the auto-focus aims mistakenly at the background (which will appear sharp) instead of the foreground, where usually the object of interest is. In this paper we propose an automatic procedure that (i) estimates the amount of out-of-focus blur, (ii) segments the image into foreground and background incorporating clues from the blurriness, (iii) recovers the sharp foreground, and finally (iv) blurs the background to refocus the scene. On several real photographs with blurry foreground and sharp background, we demonstrate the effectiveness and limitations of our method.

Inferring latent structures from observations helps to model and possibly also understand underlying data generating processes. A rich class of latent structures is the latent trees, i.e., tree-structured distributions involving latent variables where the visible variables are leaves. These are also called hierarchical latent class (HLC) models. Zhang and Kocka proposed a search algorithm for learning such models in the spirit of Bayesian network structure learning. While such an approach can find good solutions, it can be computationally expensive. As an alternative, we investigate two greedy procedures: the BIN-G algorithm determines both the structure of the tree and the cardinality of the latent variables in a bottom-up fashion. The BIN-A algorithm first determines the tree structure using agglomerative hierarchical clustering, and then determines the cardinality of the latent variables as for BIN-G. We show that even with restricting ourselves to binary trees, we obtain HLC models of comparable quality to Zhang's solutions (in terms of cross-validated log-likelihood), while being generally faster to compute. This claim is validated by a comprehensive comparison on several data sets. Furthermore, we demonstrate that our methods are able to estimate interpretable latent structures on real-world data with a large number of variables. By applying our method to a restricted version of the 20 newsgroups data, these models turn out to be related to topic models, and on data from the PASCAL Visual Object Classes (VOC) 2007 challenge, we show how such treestructured models help us understand how objects co-occur in images. For reproducibility of all experiments in this paper, all code and data sets (or links to data) are available at http://people.kyb.tuebingen.mpg.de/harmeling/code/ltt-1.4.tar.

Ultimately being motivated by facilitating space-variant blind deconvolution, we present a class of linear transformations, that are expressive enough for space-variant filters, but at the same time especially designed for efficient matrix-vector-multiplications. Successful results on astronomical imaging through atmospheric turbulences and on noisy magnetic resonance images of constantly moving objects demonstrate the practical significance of our approach.

We formulate the multiframe blind deconvolution problem in an incremental expectation maximization (EM) framework. Beyond deconvolution, we show how to use the same framework to address: (i) super-resolution despite noise and unknown blurring; (ii) saturation-correction of overexposed pixels that confound image restoration. The abundance of data allows us to address both of these without using explicit image or blur priors. The end result is a simple but effective algorithm with no hyperparameters. We apply this algorithm to real-world images from astronomy and to super resolution tasks: for both, our algorithm yields increased resolution and deconvolved images simultaneously.

Modelling camera shake as a space-invariant convolution simplifies the problem of removing camera shake, but often insufficiently models actual motion blur such as those due to camera rotation and movements outside the sensor plane or when objects in the scene have different distances to the camera. In order to overcome such limitations we contribute threefold: (i) we introduce a taxonomy of camera shakes, (ii) we show how to combine a recently introduced framework for space-variant filtering based on overlap-add from Hirsch et al.~and a fast algorithm for single image blind deconvolution for space-invariant filters from Cho and Lee to introduce a method for blind deconvolution for space-variant blur. And (iii), we present an experimental setup for evaluation that allows us to take images with real camera shake while at the same time record the space-variant point spread function corresponding to that blur. Finally, we demonstrate that our method is able to deblur images degraded by spatially-varying blur originating from real camera shake.

We study the problem of object classification when training and test classes are disjoint, i.e. no training examples of the target classes are available. This setup has hardly been studied in computer vision research, but it is the rule rather than the exception, because the world contains tens of thousands of different object classes and for only a very few of them image, collections have been formed and annotated with suitable class labels. In this paper, we tackle the problem by introducing attribute-based classification. It performs object detection based on a human-specified high-level description of the target objects instead of training images. The description consists of arbitrary semantic attributes, like shape, color or even geographic information. Because such properties transcend the specific learning task at hand, they can be pre-learned, e.g. from image datasets unrelated to the current task. Afterwards, new classes can be detected based on their attribute representation, without the need for a new training phase. In order to evaluate our method and to facilitate research in this area, we have assembled a new large-scale dataset, “Animals with Attributes”, of over 30,000 animal images that match the 50 classes in Osherson's classic table of how strongly humans associate 85 semantic attributes with animal classes. Our experiments show that by using an attribute layer it is indeed possible to build a learning object detection system that does not require any training images of the target classes.

Abstract We introduce a system for textual entailment that is based on a probabilistic model of entailment. The model is defined using a calculus of transformations on dependency trees, which is characterized by the fact that derivations in that calculus preserve the truth only with a certain probability. The calculus is successfully evaluated on the datasets of the PASCAL Challenge on Recognizing Textual Entailment.

We propose kTDSEP, a kernel-based algorithm for nonlinear blind source separation (BSS). It combines complementary research fields: kernel feature spaces and BSS using temporal information. This yields an efficient algorithm for nonlinear BSS with invertible nonlinearity. Key assumptions are that the kernel feature space is chosen rich enough to approximate the nonlinearity and that signals of interest contain temporal information. Both assumptions are fulfilled for a wide set of real-world applications. The algorithm works as follows: First, the data are (implicitly) mapped to a high (possibly infinite)—dimensional kernel feature space. In practice, however, the data form a smaller submanifold in feature space—even smaller than the number of training data points—a fact that has already been used by, for example, reduced set techniques for support vector machines. We propose to adapt to this effective dimension as a preprocessing step and to construct an orthonormal basis of this submanifold. The latter dimension-reduction step is essential for making the subsequent application of BSS methods computationally and numerically tractable. In the reduced space, we use a BSS algorithm that is based on second-order temporal decorrelation. Finally, we propose a selection procedure to obtain the original sources from the extracted nonlinear components automatically. Experiments demonstrate the excellent performance and efficiency of our kTDSEP algorithm for several problems of nonlinear BSS and for more than two sources.

In kernel based learning the data is mapped to a kernel feature space of a dimension that corresponds to the number of training data points. In practice, however, the data forms a smaller submanifold in feature space, a fact that has been used e.g. by reduced set techniques for SVMs. We propose a new mathematical construction that permits to adapt to the in- trinsic dimension and to find an orthonormal basis of this submanifold. In doing so, computations get much simpler and more important our theoretical framework allows to derive elegant kernelized blind source separation (BSS) algorithms for arbitrary invertible nonlinear mixings. Experiments demonstrate the good performance and high computational efficiency of our kTDSEP algorithm for the problem of nonlinear BSS.