Firas Laakom

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

Neural networks are composed of multiple layers arranged in a hierarchical structure jointly trained with a gradient-based optimization, where the errors are back-propagated from the last layer back to the first one. At each optimization step, neurons at a given layer receive feedback from neurons belonging to higher layers of the hierarchy. In this paper, we propose to complement this traditional 'between-layer' feedback with additional 'within-layer' feedback to encourage the diversity of the activations within the same layer. To this end, we measure the pairwise similarity between the outputs of the neurons and use it to model the layer's overall diversity. We present an extensive empirical study confirming that the proposed approach enhances the performance of several state-of-the-art neural network models in multiple tasks. The code is publically available at \url{https://github.com/firasl/AAAI-23-WLD-Reg}

Fahad Sohrab, Firas Laakom, Moncef Gabbouj

In this paper, we present an adaptation of Newton's method for the optimization of Subspace Support Vector Data Description (S-SVDD). The objective of S-SVDD is to map the original data to a subspace optimized for one-class classification, and the iterative optimization process of data mapping and description in S-SVDD relies on gradient descent. However, gradient descent only utilizes first-order information, which may lead to suboptimal results. To address this limitation, we leverage Newton's method to enhance data mapping and data description for an improved optimization of subspace learning-based one-class classification. By incorporating this auxiliary information, Newton's method offers a more efficient strategy for subspace learning in one-class classification as compared to gradient-based optimization. The paper discusses the limitations of gradient descent and the advantages of using Newton's method in subspace learning for one-class classification tasks. We provide both linear and nonlinear formulations of Newton's method-based optimization for S-SVDD. In our experiments, we explored both the minimization and maximization strategies of the objective. The results demonstrate that the proposed optimization strategy outperforms the gradient-based S-SVDD in most cases.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

Despite the superior performance of CNN, deploying them on low computational power devices is still limited as they are typically computationally expensive. One key cause of the high complexity is the connection between the convolution layers and the fully connected layers, which typically requires a high number of parameters. To alleviate this issue, Bag of Features (BoF) pooling has been recently proposed. BoF learns a dictionary, that is used to compile a histogram representation of the input. In this paper, we propose an approach that builds on top of BoF pooling to boost its efficiency by ensuring that the items of the learned dictionary are non-redundant. We propose an additional loss term, based on the pair-wise correlation of the items of the dictionary, which complements the standard loss to explicitly regularize the model to learn a more diverse and rich dictionary. The proposed strategy yields an efficient variant of BoF and further boosts its performance, without any additional parameters.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

Energy-based learning is a powerful learning paradigm that encapsulates various discriminative and generative approaches. An energy-based model (EBM) is typically formed of inner-model(s) that learn a combination of the different features to generate an energy mapping for each input configuration. In this paper, we focus on the diversity of the produced feature set. We extend the probably approximately correct (PAC) theory of EBMs and analyze the effect of redundancy reduction on the performance of EBMs. We derive generalization bounds for various learning contexts, i.e., regression, classification, and implicit regression, with different energy functions and we show that indeed reducing redundancy of the feature set can consistently decrease the gap between the true and empirical expectation of the energy and boosts the performance of the model.

Firas Laakom, Fahad Sohrab, Jenni Raitoharju et al.

One-class classification refers to approaches of learning using data from a single class only. In this paper, we propose a deep learning one-class classification method suitable for multimodal data, which relies on two convolutional autoencoders jointly trained to reconstruct the positive input data while obtaining the data representations in the latent space as compact as possible. During inference, the distance of the latent representation of an input to the origin can be used as an anomaly score. Experimental results using a multimodal macroinvertebrate image classification dataset show that the proposed multimodal method yields better results as compared to the unimodal approach. Furthermore, study the effect of different input image sizes, and we investigate how recently proposed feature diversity regularizers affect the performance of our approach. We show that such regularizers improve performance.

Alfonso Amayuelas, Firas Laakom, Piotr Piękos et al.

The use of LLMs for code generation has naturally extended to code testing and evaluation. As codebases grow in size and complexity, so does the need for automated test generation. Current approaches for LLM-based test generation rely on strategies that maximize immediate coverage gain, a greedy approach that plateaus on code where reaching deep branches requires setup steps that individually yield zero new coverage. Drawing on principles of Bayesian exploration, we treat the program's branch structure as an unknown environment, and an evolving coverage map as a proxy probabilistic posterior representing what the LLM has discovered so far. Our method, CovQValue, feeds the coverage map back to the LLM, generates diverse candidate plans in parallel, and selects the most informative plan by LLM-estimated Q-values, seeking actions that balance immediate branch discovery with future reachability. Our method outperforms greedy selection on TestGenEval Lite, achieving 51-77% higher branch coverage across three popular LLMs and winning on 77-84% of targets. In addition, we build a benchmark for iterative test generation, RepoExploreBench, where they achieve 40-74%. These results show the potential of curiosity-driven planning methods for LLM-based exploration, enabling more effective discovery of program behavior through sequential interaction

Wenyi Wang, Piotr Piękos, Li Nanbo et al.

Recent studies operationalize self-improvement through coding agents that edit their own codebases. They grow a tree of self-modifications through expansion strategies that favor higher software engineering benchmark performance, assuming that this implies more promising subsequent self-modifications. However, we identify a mismatch between the agent's self-improvement potential (metaproductivity) and its coding benchmark performance, namely the Metaproductivity-Performance Mismatch. Inspired by Huxley's concept of clade, we propose a metric ($\mathrm{CMP}$) that aggregates the benchmark performances of the descendants of an agent as an indicator of its potential for self-improvement. We show that, in our self-improving coding agent development setting, access to the true $\mathrm{CMP}$ is sufficient to simulate how the Gödel Machine would behave under certain assumptions. We introduce the Huxley-Gödel Machine (HGM), which, by estimating $\mathrm{CMP}$ and using it as guidance, searches the tree of self-modifications. On SWE-bench Verified and Polyglot, HGM outperforms prior self-improving coding agent development methods while using fewer allocated CPU hours. Last but not least, HGM demonstrates strong transfer to other coding datasets and large language models. The agent optimized by HGM on SWE-bench Verified with GPT-5-mini and evaluated on SWE-bench Lite with GPT-5 achieves human-level performance, matching the best officially checked results of human-engineered coding agents. Our code is publicly available at https://github.com/metauto-ai/HGM.

Yimeng Chen, Piotr Piȩkos, Mateusz Ostaszewski et al.

Evaluating the scientific discovery capabilities of large language model based agents, particularly how they cope with varying environmental complexity and utilize prior knowledge, requires specialized benchmarks currently lacking in the landscape. To address this gap, we introduce \textsc{PhysGym}, a novel benchmark suite and simulation platform for rigorously assessing LLM-based scientific reasoning in interactive physics environments. \textsc{PhysGym}'s primary contribution lies in its sophisticated control over the level of prior knowledge provided to the agent. This allows researchers to dissect agent performance along axes including the complexity of the problem and the prior knowledge levels. The benchmark comprises a suite of interactive simulations, where agents must actively probe environments, gather data sequentially under constraints and formulate hypotheses about underlying physical laws. \textsc{PhysGym} provides standardized evaluation protocols and metrics for assessing hypothesis accuracy and model fidelity. We demonstrate the benchmark's utility by presenting results from baseline LLMs, showcasing its ability to differentiate capabilities based on varying priors and task complexity.

Umut Cem Entok, Firas Laakom, Farhad Pakdaman et al.

Most scenes are illuminated by several light sources, where the traditional assumption of uniform illumination is invalid. This issue is ignored in most color constancy methods, primarily due to the complex spatial impact of multiple light sources on the image. Moreover, most existing multi-illuminant methods fail to preserve the smooth change of illumination, which stems from spatial dependencies in natural images. Motivated by this, we propose a novel multi-illuminant color constancy method, by learning pixel-wise illumination maps caused by multiple light sources. The proposed method enforces smoothness within neighboring pixels, by regularizing the training with the total variation loss. Moreover, a bilateral filter is provisioned further to enhance the natural appearance of the estimated images, while preserving the edges. Additionally, we propose a label-smoothing technique that enables the model to generalize well despite the uncertainties in ground truth. Quantitative and qualitative experiments demonstrate that the proposed method outperforms the state-of-the-art.

Firas Laakom, Yuheng Bu, Moncef Gabbouj

Existing generalization theories of supervised learning typically take a holistic approach and provide bounds for the expected generalization over the whole data distribution, which implicitly assumes that the model generalizes similarly for all the classes. In practice, however, there are significant variations in generalization performance among different classes, which cannot be captured by the existing generalization bounds. In this work, we tackle this problem by theoretically studying the class-generalization error, which quantifies the generalization performance of each individual class. We derive a novel information-theoretic bound for class-generalization error using the KL divergence, and we further obtain several tighter bounds using the conditional mutual information (CMI), which are significantly easier to estimate in practice. We empirically validate our proposed bounds in different neural networks and show that they accurately capture the complex class-generalization error behavior. Moreover, we show that the theoretical tools developed in this paper can be applied in several applications beyond this context.

Firas Laakom, Haobo Chen, Jürgen Schmidhuber et al.

Despite substantial progress in promoting fairness in high-stake applications using machine learning models, existing methods often modify the training process, such as through regularizers or other interventions, but lack formal guarantees that fairness achieved during training will generalize to unseen data. Although overfitting with respect to prediction performance has been extensively studied, overfitting in terms of fairness loss has received far less attention. This paper proposes a theoretical framework for analyzing fairness generalization error through an information-theoretic lens. Our novel bounding technique is based on Efron-Stein inequality, which allows us to derive tight information-theoretic fairness generalization bounds with both Mutual Information (MI) and Conditional Mutual Information (CMI). Our empirical results validate the tightness and practical relevance of these bounds across diverse fairness-aware learning algorithms. Our framework offers valuable insights to guide the design of algorithms improving fairness generalization.

Li Nanbo, Firas Laakom, Yucheng Xu et al.

World modelling is essential for understanding and predicting the dynamics of complex systems by learning both spatial and temporal dependencies. However, current frameworks, such as Transformers and selective state-space models like Mambas, exhibit limitations in efficiently encoding spatial and temporal structures, particularly in scenarios requiring long-term high-dimensional sequence modelling. To address these issues, we propose a novel recurrent framework, the \textbf{FACT}ored \textbf{S}tate-space (\textbf{FACTS}) model, for spatial-temporal world modelling. The FACTS framework constructs a graph-structured memory with a routing mechanism that learns permutable memory representations, ensuring invariance to input permutations while adapting through selective state-space propagation. Furthermore, FACTS supports parallel computation of high-dimensional sequences. We empirically evaluate FACTS across diverse tasks, including multivariate time series forecasting, object-centric world modelling, and spatial-temporal graph prediction, demonstrating that it consistently outperforms or matches specialised state-of-the-art models, despite its general-purpose world modelling design.

Tyler A. Chang, Catherine Arnett, Abdelrahman Eldesokey et al. · uw

To date, there exist almost no culturally-specific evaluation benchmarks for large language models (LLMs) that cover a large number of languages and cultures. In this paper, we present Global PIQA, a participatory commonsense reasoning benchmark for over 100 languages, constructed by hand by 335 researchers from 65 countries around the world. The 116 language varieties in Global PIQA cover five continents, 14 language families, and 23 writing systems. In the non-parallel split of Global PIQA, over 50% of examples reference local foods, customs, traditions, or other culturally-specific elements. We find that state-of-the-art LLMs perform well on Global PIQA in aggregate, but they exhibit weaker performance in lower-resource languages (up to a 37% accuracy gap, despite random chance at 50%). Open models generally perform worse than proprietary models. Global PIQA highlights that in many languages and cultures, everyday knowledge remains an area for improvement, alongside more widely-discussed capabilities such as complex reasoning and expert knowledge. Beyond its uses for LLM evaluation, we hope that Global PIQA provides a glimpse into the wide diversity of cultures in which human language is embedded.

Firas Laakom, Jenni Raitoharju, Nikolaos Passalis et al.

In this paper, we consider the problem of non-linear dimensionality reduction under uncertainty, both from a theoretical and algorithmic perspectives. Since real-world data usually contain measurements with uncertainties and artifacts, the input space in the proposed framework consists of probability distributions to model the uncertainties associated with each sample. We propose a new dimensionality reduction framework, called NGEU, which leverages uncertainty information and directly extends several traditional approaches, e.g., KPCA, MDA/KMFA, to receive as inputs the probability distributions instead of the original data. We show that the proposed NGEU formulation exhibits a global closed-form solution, and we analyze, based on the Rademacher complexity, how the underlying uncertainties theoretically affect the generalization ability of the framework. Empirical results on different datasets show the effectiveness of the proposed framework.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

Autoencoders are a type of unsupervised neural networks, which can be used to solve various tasks, e.g., dimensionality reduction, image compression, and image denoising. An AE has two goals: (i) compress the original input to a low-dimensional space at the bottleneck of the network topology using an encoder, (ii) reconstruct the input from the representation at the bottleneck using a decoder. Both encoder and decoder are optimized jointly by minimizing a distortion-based loss which implicitly forces the model to keep only those variations of input data that are required to reconstruct the and to reduce redundancies. In this paper, we propose a scheme to explicitly penalize feature redundancies in the bottleneck representation. To this end, we propose an additional loss term, based on the pair-wise correlation of the neurons, which complements the standard reconstruction loss forcing the encoder to learn a more diverse and richer representation of the input. We tested our approach across different tasks: dimensionality reduction using three different dataset, image compression using the MNIST dataset, and image denoising using fashion MNIST. The experimental results show that the proposed loss leads consistently to superior performance compared to the standard AE loss.

Firas Laakom, Kateryna Chumachenko, Jenni Raitoharju et al.

Recently, there has been an increasing interest in applying attention mechanisms in Convolutional Neural Networks (CNNs) to solve computer vision tasks. Most of these methods learn to explicitly identify and highlight relevant parts of the scene and pass the attended image to further layers of the network. In this paper, we argue that such an approach might not be optimal. Arguably, explicitly learning which parts of the image are relevant is typically harder than learning which parts of the image are less relevant and, thus, should be ignored. In fact, in vision domain, there are many easy-to-identify patterns of irrelevant features. For example, image regions close to the borders are less likely to contain useful information for a classification task. Based on this idea, we propose to reformulate the attention mechanism in CNNs to learn to ignore instead of learning to attend. Specifically, we propose to explicitly learn irrelevant information in the scene and suppress it in the produced representation, keeping only important attributes. This implicit attention scheme can be incorporated into any existing attention mechanism. In this work, we validate this idea using two recent attention methods Squeeze and Excitation (SE) block and Convolutional Block Attention Module (CBAM). Experimental results on different datasets and model architectures show that learning to ignore, i.e., implicit attention, yields superior performance compared to the standard approaches.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

We study the diversity of the features learned by a two-layer neural network trained with the least squares loss. We measure the diversity by the average $L_2$-distance between the hidden-layer features and theoretically investigate how learning non-redundant distinct features affects the performance of the network. To do so, we derive novel generalization bounds depending on feature diversity based on Rademacher complexity for such networks. Our analysis proves that more distinct features at the network's units within the hidden layer lead to better generalization. We also show how to extend our results to deeper networks and different losses.

Firas Laakom, Jenni Raitoharju, Nikolaos Passalis et al.

spectral-based subspace learning is a common data preprocessing step in many machine learning pipelines. The main aim is to learn a meaningful low dimensional embedding of the data. However, most subspace learning methods do not take into consideration possible measurement inaccuracies or artifacts that can lead to data with high uncertainty. Thus, learning directly from raw data can be misleading and can negatively impact the accuracy. In this paper, we propose to model artifacts in training data using probability distributions; each data point is represented by a Gaussian distribution centered at the original data point and having a variance modeling its uncertainty. We reformulate the Graph Embedding framework to make it suitable for learning from distributions and we study as special cases the Linear Discriminant Analysis and the Marginal Fisher Analysis techniques. Furthermore, we propose two schemes for modeling data uncertainty based on pair-wise distances in an unsupervised and a supervised contexts.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

Computational color constancy is a preprocessing step used in many camera systems. The main aim is to discount the effect of the illumination on the colors in the scene and restore the original colors of the objects. Recently, several deep learning-based approaches have been proposed to solve this problem and they often led to state-of-the-art performance in terms of average errors. However, for extreme samples, these methods fail and lead to high errors. In this paper, we address this limitation by proposing to aggregate different deep learning methods according to their output uncertainty. We estimate the relative uncertainty of each approach using Monte Carlo dropout and the final illumination estimate is obtained as the sum of the different model estimates weighted by the log-inverse of their corresponding uncertainties. The proposed framework leads to state-of-the-art performance on INTEL-TAU dataset.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

In this paper, we propose a novel unsupervised color constancy method, called Probabilistic Color Constancy (PCC). We define a framework for estimating the illumination of a scene by weighting the contribution of different image regions using a graph-based representation of the image. To estimate the weight of each (super-)pixel, we rely on two assumptions: (Super-)pixels with similar colors contribute similarly and darker (super-)pixels contribute less. The resulting system has one global optimum solution. The proposed method achieves competitive performance, compared to the state-of-the-art, on INTEL-TAU dataset.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

In this paper, we describe a new large dataset for illumination estimation. This dataset, called INTEL-TAU, contains 7022 images in total, which makes it the largest available high-resolution dataset for illumination estimation research. The variety of scenes captured using three different camera models, namely Canon 5DSR, Nikon D810, and Sony IMX135, makes the dataset appropriate for evaluating the camera and scene invariance of the different illumination estimation techniques. Privacy masking is done for sensitive information, e.g., faces. Thus, the dataset is coherent with the new General Data Protection Regulation (GDPR). Furthermore, the effect of color shading for mobile images can be evaluated with INTEL-TAU dataset, as both corrected and uncorrected versions of the raw data are provided. Furthermore, this paper benchmarks several color constancy approaches on the proposed dataset.

Firas Laakom, Nikolaos Passalis, Jenni Raitoharju et al.

In this paper, we propose a novel color constancy approach, called Bag of Color Features (BoCF), building upon Bag-of-Features pooling. The proposed method substantially reduces the number of parameters needed for illumination estimation. At the same time, the proposed method is consistent with the color constancy assumption stating that global spatial information is not relevant for illumination estimation and local information ( edges, etc.) is sufficient. Furthermore, BoCF is consistent with color constancy statistical approaches and can be interpreted as a learning-based generalization of many statistical approaches. To further improve the illumination estimation accuracy, we propose a novel attention mechanism for the BoCF model with two variants based on self-attention. BoCF approach and its variants achieve competitive, compared to the state of the art, results while requiring much fewer parameters on three benchmark datasets: ColorChecker RECommended, INTEL-TUT version 2, and NUS8.

Firas Laakom, Jenni Raitoharju, Alexandros Iosifidis et al.

In this paper, we study the importance of pre-training for the generalization capability in the color constancy problem. We propose two novel approaches based on convolutional autoencoders: an unsupervised pre-training algorithm using a fine-tuned encoder and a semi-supervised pre-training algorithm using a novel composite-loss function. This enables us to solve the data scarcity problem and achieve competitive, to the state-of-the-art, results while requiring much fewer parameters on ColorChecker RECommended dataset. We further study the over-fitting phenomenon on the recently introduced version of INTEL-TUT Dataset for Camera Invariant Color Constancy Research, which has both field and non-field scenes acquired by three different camera models.